x
Optical/IP

MPLS Vendors Demo Fast Reroute

Yesterday, Isocore, an independent testing facility in McLean, Va , hosted its third public interoperability demonstration where it tested for the first time a new Internet Engineering Task Force (IETF) draft of MPLS Fast Reroute, a technology that brings Sonet-like protection to Multiprotocol Label Switching (MPLS) networks (see MPLS Fast Reroute Gains Momentum).

Cisco Systems Inc. (Nasdaq: CSCO), Avici Systems Inc. (Nasdaq: AVCI; Frankfurt: BVC7) and Juniper Networks Inc. (Nasdaq: JNPR) participated in the test using specifications from a draft written by George Swallow of Cisco, which combine work from two other drafts on Fast Reroute. The test focuses on link protection and did not include node protection mechanisms. Cisco and Avici had tested an earlier Fast Reroute draft for link protection in August.

Ixia (Nasdaq: XXIA), Spirent plc (NYSE: SPM; London: SPT), and NetTest provided the testing equipment.

The importance of the demo is twofold. For one, this is the first time that all three core routing vendors have tested this feature together. And second, it is an important step forward in developing this critical feature for MPLS.

“The idea behind doing these demos is to show certain technologies working today,” says Bijan Jabbari, president of Isocore. “But we need to work further to make sure all the features are standardized. And I think that this process will help facilitate that.” The test bed included IP core routers from three vendors: Avici’s TSR, two Cisco GSR 12404s, and Juniper’s M10. The way it was set up was that traffic entered the network at the CoSine Communications Inc. (Nasdaq: COSN) IPSX3500, which was used as the access router. It routed traffic to the Cisco GSR. The Cisco GSR then routed it to the Juniper M10, which routed it onto another Cisco GSR. When the cable was disconnected between the Cisco GSR and the Juniper M10 router, the Fast Reroute mechanism kicked in and rerouted the traffic to the Avici TSR, which was connected to the GSR and M10. Is that clear? The network restoration occurred in roughly 5 milliseconds, well within the current Sonet standard that requires 50 milliseconds restoration. Restoration in the demo has consistently run within 10 milliseconds, says Jabbari.

The reason the restoration is so quick is because the traffic is rerouted locally. There is no signaling protocol that propagates through the network to let the next hop know packets are coming, making the failover time to an alternate path extremely fast, says Don Troshynski, senior systems engineer for Avici.

The demonstration itself was not terribly exciting to watch. In fact, most of the observers didn’t even realize the test was going on until it was over. Even Tony De La Rosa, product manager for Ixia, who was running the demonstration and explaining what was happening, hadn’t realized that the line between the routers had been disconnected. Interestingly enough, this seems to be the point of demonstrations showing off reconvergence technologies. The point is to show restoration occurring so quickly that there is no noticeable loss of service.

“Restoration time can fluctuate within a certain range,” says Amrit Hanspal, product manager for MPLS and QOS at Cisco. “As long as you keep it below 50 milliseconds, it’s alright. And even then some applications will be fine if the restoration time is longer. Call signaling is usually dropped after 1 or 2 seconds.”

The vendors involved in setting up the network and the test have been working together on this project for the past few months, dedicating two full days to setting up the demonstration prior to its public debut. The work that was done to get this implemented will go a long way in helping these vendors further define the standard, says Loa Andersson, chief architect for Utfors AB.

“From an IETF perspective this is a pretty good basis for developing the protocol design,” he says. “It’s still early days for this feature, and this a good way to work out the problems early on.”

Andersson says that from a service provider perspective, this demonstration could highlight problems that he would look for when implementing the feature in his company’s own test bed.

Heidi Tse, a backbone engineer for WorldCom Inc. (OTC: WCOEQ), says that these demonstrations are important to a degree, but she emphasizes that carriers must still test the gear themselves.

“We don’t want to be a vendor’s guinea pig,” she says. “But you also have to understand that this is not a sophisticated test, and we would have to test it much more in depth ourselves if we were going to implement this feature. Still, tests like this are important because it will help move the standard along and we all need the standard before we do anything.”

Isocore also demonstrated yesterday interoperability for label switched path (LSP) protection using gear from Marconi plc (Nasdaq/London: MONI). And it demonstrated provisioning of Layer 3 MPLS/BGP (Border Gateway Protocol) and Layer 2 Draft Martini virtual private networks using gear from Cisco, CoSine Communications Inc. (Nasdaq: COSN), Foundry Networks Inc. (Nasdaq: FDRY), Extreme Networks Inc. (Nasdaq: EXTR), Lucent Technologies Inc. (NYSE: LU), and Redback Networks Inc. (Nasdaq: RBAK).

Jabbari says Isocore plans to conduct more interoperability tests next year. Specifically, he would like to test node detour in MPLS fast reroute in addition to the link protection that was demonstrated already. But he says he must wait for all the vendors to have a working implementation of that spec before a test can be designed. Andersson and Tse will likely be tuning in again to that one.

“Node protection is on my Christmas wish list,” says Andersson. “But we need to take one thing at a time. I’m glad they’ve got the link protection working. Now they can move on to the node protection.”

— Marguerite Reardon, Senior Editor, Light Reading
www.lightreading.com
sgan201 12/4/2012 | 9:23:08 PM
re: MPLS Vendors Demo Fast Reroute Hi All,
The latest issue of Business Communications Review.. Equant is quoted as having more than 500 customers for IP VPN services.. You can draw your own conclusion from that statement..
By the way, Equant is one of the largest IP VPN provider...
fiber_r_us 12/4/2012 | 9:24:00 PM
re: MPLS Vendors Demo Fast Reroute SNA is supported, just like any other protocol. FR is agnostic to the protocols above it. If you wished to give SNA priority, you would simply use one of the higher service classes.
sgan201 12/4/2012 | 9:24:05 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
Is SNA supported or not supported on this FR over MPLS network??
sgan201 12/4/2012 | 9:24:05 PM
re: MPLS Vendors Demo Fast Reroute Hi,
It would be very interesting to find out what is the maximum trunk utilization level that the MPLS network is traffic engineered to. It would seem like it has to 40% or even 20% to achieve that kind of SLA...
sgan201 12/4/2012 | 9:24:06 PM
re: MPLS Vendors Demo Fast Reroute Worldcom FR according to Current Analysis
Frame Delivery CIR PVCs 99.99%
PVC Delay 60 ms one-way
Port Availability 100%
Data Delivery Ratio 99.99% within CIR
Network Availability 100%
Mean Time to Repair 2 hours On-Net



Back to Top

sgan201 12/4/2012 | 9:24:06 PM
re: MPLS Vendors Demo Fast Reroute Sprint FR according to Current Analysis
Frame Delivery CIR PVCs 99.90%
PVC Delay 55-130 ms
Port Availability 100% for BMAN
Data Delivery Ratio 99% w/no CIR, 99.9% w/CIR spec.
Network Availability 100% for BMAN
Mean Time to Repair ICB


sgan201 12/4/2012 | 9:24:06 PM
re: MPLS Vendors Demo Fast Reroute AT&T FR according to current Analysis
Frame Delivery CIR PVCs 99.90%
PVC Delay 60 ms one way
Port Availability 99.99%
Data Delivery Ratio 99.99% at CIR
Network Availability 99.99%
Mean Time to Repair 4hrs-8hrs if dispatch is needed



Back to Top

gardner 12/4/2012 | 9:24:09 PM
re: MPLS Vendors Demo Fast Reroute Actually if you zoom it all the way (I have 5.0 Acrobat reader) it is very readable.

It says:

*First availability figure is for single-attached access. The second is for dual-attached.
** All latency guarantees are route-specific. The range provided represents the minimum and maximum values.
***For optimized traffic, committed traffic carries a 100% delivery SLA. Burst traffic carries a 99.95% delivery SLA.

And there you have it sports fans.
sgan201 12/4/2012 | 9:24:09 PM
re: MPLS Vendors Demo Fast Reroute http://www.nwfusion.com/news/2...

Check out Worldcom Economy FR and Bundled FR pricing..
fiber_r_us 12/4/2012 | 9:24:10 PM
re: MPLS Vendors Demo Fast Reroute Zoom in on it and its readable (barely)...
sgan201 12/4/2012 | 9:24:12 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
The * part of the document is not readable...
Care to tell us the fine print...
sgan201 12/4/2012 | 9:24:13 PM
re: MPLS Vendors Demo Fast Reroute Hi LightGauge,
I was responding to a question of whether an ATM switch can do routing and IP VPN = single box solution. Now, whether RPM in MGX is a good solution or not for a particular network is another matter..

I was trying to illustrate the point that saying MPLS box is the only way to do single box/single core solution is not true..
lob 12/4/2012 | 9:24:14 PM
re: MPLS Vendors Demo Fast Reroute > if you have the choice of bottled water and tap
> water and they both costs the same to you.

I'd see it as a business opportunity (because, obviously, there's a disbalance) or as a reason to write to my elected representative in regard to clear evidence of market manipulation. In case of telecom, both :)
LightGaugeGuitarString 12/4/2012 | 9:24:15 PM
re: MPLS Vendors Demo Fast Reroute Hi Dreamer,

As a matter of fact, I do indeed want to make money, but it will not be via manufacturing routers, or FR/ATM switches or multiservice boxes.

Yes, the Cisco 75XX/72XX routers are the widest deployed... and everyone that I talk to who has deployed them absolutely despises them. They are underpowered, have lousy Frac T1 density, ML-PPP doesn't work well, and start to fall apart as soon as you turn up a few filtering rules (unless they've finally got that stuff working on the network processor and off of the RISC).

The point that I was trying to make is that deploying traditional ATM switches with router blades like the MGX as a router is a mistake, from a performance, and, most importantly, a cost basis. And I'm not talking about deploying it as a core router, I mean deploying it as an edge router. You seem to be misinterpreting that as telling you that you should build a core router? I'm just saying that the MGX makes a lousy router.

If you're saying there's more money to be had at the edge, I don't think anybody is arguing with you. I'm certainly not. It is a crowded field, however, but I wish you the best of luck.

LGGS

------------------------------------------------
Hi LightGaugeGuitarString,
You want to scale or you want to make money as equipment manufacturer??
The best selling SP routers is Cisco 75XX/72XX..
GSR/Core Router market is much smaller than that..

I would design my box to target the largest market with the best profit margin..

Why would I design a bigger box with smaller addressable market???

fiber_r_us 12/4/2012 | 9:24:18 PM
re: MPLS Vendors Demo Fast Reroute
This is the only SLA I could find relating to jitter for frame relay or Layer 2 MPLS services was this one:

http://www.level3.com/userimag...

Couldn't find anything equivalent by searching Sprint or Worldcom's sites... Anybody have any direct links?
fiber_r_us 12/4/2012 | 9:24:19 PM
re: MPLS Vendors Demo Fast Reroute
>Do you know that Cisco 75XX/72XX revenue is many >times larger than GSR 12XXX and 10XXX??

And the 65xx revenue is larger than all of them put together... Maybe we should all make our networks with Ethernet switches since they generate the most revenue for Cisco! I am at a loss as to what the breakdown of product revenue for a hardware vendor has to do with our discussion of the most cost-effective network architecture for service providers.
sgan201 12/4/2012 | 9:24:20 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
If you interested in the truth, you can check out a few web site and compare the SLA of IP VPN versus FR. If not, it does not matter to me.. I am done trying to convince you. The truth will come out anyhow..
In fact, the more people wasting their time on IP/MPLS and IP VPN, the better off I am..
sgan201 12/4/2012 | 9:24:20 PM
re: MPLS Vendors Demo Fast Reroute Hi LightGaugeGuitarString,
You want to scale or you want to make money as equipment manufacturer??
The best selling SP routers is Cisco 75XX/72XX..
GSR/Core Router market is much smaller than that..

I would design my box to target the largest market with the best profit margin..

Why would I design a bigger box with smaller addressable market???

sgan201 12/4/2012 | 9:24:20 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
Do you know that Cisco 75XX/72XX revenue is many times larger than GSR 12XXX and 10XXX??
sgan201 12/4/2012 | 9:24:20 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
IP backbone is not where the money is..
Frankly speaking, ATM Switch folks do not care about that..
Most access to the IP backbone is via FR/ATM..
That is where you make big bucks..
And, FR network/backbone is where the money is..
If not, why all those router is working hard to support FR via draft-Martini. They could do only IP..

For every OC-48, you need a lot of DS1 ports..
I will let Cisco/Juniper sell the core routers to everyone. If I win the edge aggregation piece, I will make more money..
sgan201 12/4/2012 | 9:24:21 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
Why enter into all these stupid core router tests when you have the revenue and you are laughing your way to the bank?? Better keep quiet and hope nobody notice and no competition.
Look, how many next-gen ATM switches is out there versus next-gen IP/MPLS boxes..
If you have customer and revenue, keeping quiet mode is the best way..
sgan201 12/4/2012 | 9:24:22 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
Nortel Passport is an LER/Edge router..
Core router market is much smaller than edge router.. Edge router market is 2X Core router market..
LightGaugeGuitarString 12/4/2012 | 9:24:26 PM
re: MPLS Vendors Demo Fast Reroute Fiber-r-us dais:
"Just because a box supports IP routing protcols, does not make it an IP router."

I have to echo fiber-r-us' sentiment. Unless a product has been designed and architected to perform as a router, it won't scale. "Offboard" routing engines (as they're called) on ATM switches traditionally end up having centralized route processing bottlenecks (FIB, RIB, exception handling, IGP, BGP, etc. all done on one centralized processor on the blade). Sure you can add another routing blade (you can do this on the MGX) but this usually comes with a restriction - the routing blade only applies to a subset of the slots (this is true for the MGX in at least one configuration that I'm aware of. It was also true for the Nortel PP15K at one point in its history, not sure if it's still the case). So what you've really done is logically partitioned an ATM switch into separate low-powered routing entities.

Clearly, when you look at "state of the art" in core IP routing/switching design, you don't see that on the MGX.

If I draw an analogy, it's just like you can't buy a Cisco GSR or Juniper M- or T-series router, fill it up with ATM blades, and call it an ATM switch, despite what the marketing folk will tell you :)

Anyways, there seems to be two competing views/visions here (I apologize in advance for points I've missed or misrepresentations):

Dreamer's Vision:
Service adaptation at the customer prem or SP edge into FR or ATM, and an ATM core with integrated routing in the ATM sitches, or one-armed routers, or a separate "Internet" IP network altogether. He acknowledges that the "Internet" exists and must exist, but contends that for value-added services (read: high margin) FR and ATM is a better choice than MPLS due to its imbedded base, and that ATM can encapsulate other services (FR, ATM, Ethernet, voice, etc.) effectively. PNNI and AINI has resulted in effective dynamic bandwidth management facilitating operations. SVC capacity has been a problem in the past, but these days new ATM switches have vastly improved SVC capacity (just as core routers have vastly improved their packet forwarding and route processing). Dreamer contends that FR & ATM represents the majority of carrier revenue, and ultimately this *revenue* momentum drive carriers to adopt ATM as the underlying layer2 mechansim. Spending R&D on MPLS is a wasted effort, as everything that MPLS is trying to be you can get with ATM (or ATM+SONET, when discussing Fast Reroute)

fiber-r-us' vision:
Service adaptation at the customer prem or SP edge (including FR, ATM, Ethernet, etc.) into MPLS, and an IP-MPLS core with integrated routing in the MPLS LSRs, with no need for physically separate Internet devices for networks (although due to legacy IP networks incapable of MPLS, there will be some Best Effort Internet islands incapable of MPLS, but that's easy to engineer around). He acknowledges that the ATM & FR services exist and must continue to exist, but contends that for value-added services (read: high margin) MPLS is a better choice than ATM, as the consolidation of Internet plus MPLS' ability to encapsulate other services (FR, ATM, Ethernet, voice, etc.). IP Router vendors have introduced QoS capabilities that provide adequate IP QoS for a majority of the services, if not all. It is the *traffic volume* momentum of IP and MPLS' inherent synnergy with IP that is the driver - IP networks carry more volume than ATM, consolidating services to a technology that can leverage the existing imbedded IP base will result in economy of scale gains and operations savings.

Personally, my business relies on neither, in fact is enhanced if the world is equally split on ATM versus IP-MPLS.

LGGS
fiber_r_us 12/4/2012 | 9:24:28 PM
re: MPLS Vendors Demo Fast Reroute Dreamer said:

>1) Is people willing to pay more for IP VPN compare to FR??
>Please noted that SLA for IP VPN is worst than
>FR in term of jitter and uptime..

Point me to a major provider's SLA on jitter/uptime of FR.

>Please noted that according to Infornetics, 60%
>of the IP VPN is DIY (do-it-yourself).

The discussion is on SP-operated VPNs over MPLS. DIY IP VPNs is a totally different topic that is not relavent to the discussion.

>2) If you design a network, does it costs more or less
>to provide FR service versus IP VPN services?

If I designed a network that *only* support FR/ATM over MPLS, then this would be more expensive that if I would have done FR/ATM with traditional gear. But, that has not what I have been suggesting. The idea is to run all of your data and IP services over the same MPLS infrastructure (not just FR and ATM). This would include support for: FR, ATM, Ethernet, VPLS, L3VPNs, Internet, packetized voice, packetized video, etc. When combined, the infrastructure would result in a cheaper service offering for all services as compared to offering each over its own dedicated infrastructure. This savings results not only from the sharing of hardware infrastructure, but in the massive simplification of operational issues: people, management systems, OSS systems, power, space, sparing, etc.


fiber_r_us 12/4/2012 | 9:24:29 PM
re: MPLS Vendors Demo Fast Reroute >Is the RPM inside the MGX box??
>Yes..

>Is this a single box solution??
>Yes..

Is the RPM a useful Internet router that can support GigE, OC-12, OC-48, OC-192, or 10GE?

No.

Do I know of any of the major Internet backbones that would even consider this a workable solution to build thier network?

No.

Dreamer, you can't possibly believe that you could build and support any reasonably sized Internet backbone on 7200-class gear!
fiber_r_us 12/4/2012 | 9:24:30 PM
re: MPLS Vendors Demo Fast Reroute Dreamer,

Just because a box supports IP routing protcols, does not make it an IP router. A box must be able to forward packets based on IP headers, decrement TTL, fragment, support numerous external BGP peers with 100,000s of BGP routes, etc. The Nortel web page is very unclear whether the 20000 can do this. If it can, why wasn't it entered into any core router tests? Is it supposed to compete/replace GSRs and M160s/T640s? It is not even marketed as a core Internet router.

The web page leads me to believe that the 20000 is attempting to be an MPLS LSR (core or edge) and that it could *not* be used as an Internet router. If this is true, it could potentially be a useful box as a pure LSR in the core or potentially an MPLS edge box that doesn't support the Internet. But, if that was all it was good for, why wouldn't I use GSRs or Junipers?
netskeptic 12/4/2012 | 9:24:30 PM
re: MPLS Vendors Demo Fast Reroute > why are carriers (apparently) asking for MPLS,
> and buying GSR's, M160's etc?

1. Carriers are asking for MPLS because it is essentially free, so they are just making sure that they are loading all possible free stuff with their paid equipment.

2. Carriers are buying routers because there is still (always will be) the need to route IP traffic.

Thanks,

Netskeptic

MP_UK 12/4/2012 | 9:24:32 PM
re: MPLS Vendors Demo Fast Reroute Hi dreamer,
I'm sure Cisco will hedge their bets, seems like the sensible thing to do. I can't really comment on what fiber_r_us said about the MGX routing engine being a bit weedy, but the only place I've ever seen one deployed is right next to a pair of 7500's doing all the layer 3 stuff.

I don't know enough about MPLS to give it a hard time, I don't work for a carrier, or on core technologies. However in my limited understanding, for load balancing in the core MPLS seems like quite an elegant solution, and FRR may be part of that. As for using it for VPN's or providing QoS guarantees, I am yet to be convinced...

Back to my origional post though, if the passport 15000 does everything it says it can, why are carriers (apparently) asking for MPLS, and buying GSR's, M160's etc? After all, if the end goal is a converged network (i.e. one type of box supports all services,) why is it better to use MPLS routers than IP routing ATM switches - the cost of connecting two OC-192 PoS ports is no different than connecting two oc-192 ATM ports.
sgan201 12/4/2012 | 9:24:33 PM
re: MPLS Vendors Demo Fast Reroute Hi MP_UK,
Cisco is a smart company.. Their MGX can do the same thing. They are hedging their bet..
Fundamentally, Cisco is a sales driven company.. In spite whatever they say, they always build whatever customer willing to buy.. Smart!!
They master the act of saying something but do something else..

They are also coming out OC-192c ATM interface on MGX box..
sgan201 12/4/2012 | 9:24:34 PM
re: MPLS Vendors Demo Fast Reroute Hi MP_UK,
That is my point precisely...
sgan201 12/4/2012 | 9:24:34 PM
re: MPLS Vendors Demo Fast Reroute Hi lob,
if you have the choice of bottled water and tap water and they both costs the same to you.
What would you choose??
This is the situation of IP VPN versus FR service looking from the standpoint of Enterprise Network Engineer..
nenene66 12/4/2012 | 9:24:36 PM
re: MPLS Vendors Demo Fast Reroute The answer is easy. They (Cisco and Juniper) have almost 0 market share in the ATM switch market. They would love to be able to MPLS enable carriers existing ATM networks to continue to grow their revenue. You need market share to do this.

The REAL MPLS networks will be built on top of the existing ATM/FR networks and equipment. It will not be built as a parrellel network with M160 and GSR routers (notice I said ROUTER not SWITCH) Carriers will NOT throw away their currnet ATM investment to deliver the same VPN service they deliver today with FR/ATM.

Ne
MP_UK 12/4/2012 | 9:24:37 PM
re: MPLS Vendors Demo Fast Reroute
I had a quick look at the spec sheet for the Passport 15000. So assuming it does everything that it says it can, isn't this - or any similar product - already doing what MPLS enabled core routers are trying to do? It seems to have the best of both worlds, ATM QoS reatures, and full featured IP routing in one box, all using proven technology - so that gives a converged network... right?

I'm sure there must be more to the story, but if it's possible to build a decent router and a decent ATM switch into the same box, why bother with a new protocol (MPLS)? Why aren't Cisco and Juniper building these? Just because of the cell tax...?
AAL5 12/4/2012 | 9:24:38 PM
re: MPLS Vendors Demo Fast Reroute >>>You don't need MPLS to tunnel a packer or a frame
>>>across an IP cloud. GRE or L2TP as you suggested
>>>can do it really well with no additional state
>>>or new sw in the core.
>
>> This is really obvious I donG«÷t know why you are
>> trying to compare GRE/L2TP to MPLS L3VPNs.

>You're missing the point. Your argument was that
>MPLS allows SPs to provide L3VPNs. I'm saying
>they don't need MPLS for L3VPNs. One can even
>have 2547 without MPLS.

You originally suggested GRE/L2TP as a substitute for L3VPNs,
you did not say anything about using GRE/L2TP as the tunneling
protocol and inventing something similar to BGP MPLS VPNs
to do the signaling. As mentioned previously I am not religious
about MPLS, but Juniper/Cisco have L3VPN implementations
based on MPLS so I doubt others will succeed. L3VPN based on MPLS
are available here, now and today, multi-vendor support for an IP based solution
is not going to happen soon.

>> MPLS L3VPNs provide more that tunneling; they
>> provides virtual private forwarding table services
>> that can be used to create VPNs that forward
>> information based on L3 information. I donG«÷t see
>> what your argument is here, they are NOT the same.

>VRs and VRFs are completely orthogonal to the
>tunneling techniques. Your logic is L3VPNs ==
>2547 == MPLS.

No I never said this, I was merely pointing out that GRE/L2TP are not substitutes
for BGP MPLS VPNs, they provide tunneling services only.

>> It is technically possible to provide L3VPNs based
>> on IP tunneling protocols. IG«÷m not religious about
>> MPLS, just being realistic; L3VPNs based on MPLS are
>> deployed in customer networks already by Cisco/Juniper,
>> I do not see L3VPNs based on IP succeeding.

>Thought it works both ways, what vendors do finally
>depends on what SPs ask for. I do hear some SPs saying
>they need an non-MPLS non-BGP VPN solution. I have
>to admit they are in minority. I do hope, however,
>that the rest of them do not believe VPNs == MPLS.

Yip, I know one big vendor that doesnG«÷t want MPLS in their network.
I initially was pointing out some the benefits you can get from MPLS based technologies today,
I never said that L2/L3 VPNs were not possible using IP.

>>>>> - Allows SPs to provide L2VPN services to
>>>>why not GRE?
>>> You really mean L2TPv3 don't you?
>>L2TP will work as well.
> Yes L2TP is equivalent but not GRE as you originally said.

>You're probably mixing signaling and tunneling together.
>One could perfectly have L2VPNs with GRE.
>We're diverging from the original point though,
>which is again: you don't need MPLS for L2VPNs.

Your original response to my post was trying to compare IP tunneling protocols to signaling/tunneling solutions using MPLS, this was not a correct comparison. It was only in later responses do you mentioned about the feasibility of implementing L3VPNs on IP. Your G«ˇpointG«÷ seems to get modified with every post you make.


>I wouldn't argue that MPLS FRR is not valuable. I do
>think, however, that we should have FRR or fast convergence
>in regular IP routing. This is doable and will help
>to avoid MPLS complexity.

What complexity are you talking about? I suppose you are referring to all the complexity of BGP MPLS VPNs and FRR, which you are proposing to replace with another G«ˇcomplexG«÷ solution based on IP? ThereG«÷s nothing complex about it at all; unfamiliar to some IP-only die-hards, but complex no. When you find IP-only solution implemented to replace all the functionality of an MPLS enabled core gives you we can make the comparison again.


>And the equivalent of the Internet end-to-end connectivity in MPLS is...?

That would be IPv4/IPv6 or any other L2/L3 protocol that you want to transport in an MPLS enabled network. MPLS is not the only technically possible solution, but itG«÷s the only one that can provide a solution to unify the core at OC-48 and OC-192 speeds.

You ever tried doing L2TP de-aggregation on an OC-48 or OC-192 based line card at line rate?

AAL5
lob 12/4/2012 | 9:24:41 PM
re: MPLS Vendors Demo Fast Reroute > If you look at IP traffic, 40% of the packets
> are 40 bytes (TCP acks). If you add 28 bytes,
> then you're at about 40% overhead. Most people
> found that objectionable and petitioned for a
> way to do tunneling (and source routing) with
> lower overhead.

Regrettably, most people cannot be persuaded to do some back-of-envelope calculations before voicing strong opinions :)

The mean packet size as observed at NASA Ames Internet Exchange is 420 bytes. Source: http://www.caida.org/analysis/...

Only about 8% of bandwidth is consumed by packets smaller than 200 bytes. So doubling that bandwidth does not produce much overhead.

In fact the real overhead with 28-byte GRE headers would be about 6%. A huge lot less than 20+% you get with AAL5 on the same packet mix, and there's quite a few people thinking that cell tax is quite acceptable price for doing VPNs :)

Of course, there's also an option of simply allocating a different L2 frame type to data packets with prepended 4-byte IPv4 "tunnel end" address. This lacks universality of GRE (i.e. ability to tunnel through opaque domains) but so
does MPLS.

So, the real reason why GRE remained useless is not overhead (it is tolerable) but, rather, inability of OFRV's boxes to do it fast.
lob 12/4/2012 | 9:24:41 PM
re: MPLS Vendors Demo Fast Reroute > Do you drink water from the tap or you buy
> bottled water??

I prefer bottled, but if I'm thirsty and there's no bottled water around I'd drink from tap.

What bandwith reservation "QoS" gives is "sorry, no bottled water, go die from thirst".

> Why should I sacrify QOS when it is the same
> price or lower than without QOS??

It is NOT the same price; you pay arm and leg in
costs of overly complicated boxes and operational support of those.

You are mixing "premium service" and "guaranteed service".
Tony Li 12/4/2012 | 9:24:42 PM
re: MPLS Vendors Demo Fast Reroute From what I remember, it was ignored because
how certain vendors implemented it in their
forwarding. That it was all done on the slow
path meaning that GRE could not be used as more
than a toy. That may be fixed in newer equipment
(I dont know), but the perception is still there.
----------------

Interesting... because we would have happily made it faster, but folks told us not to bother because of the overhead.

----------------

Also, most people who do tunneling want absolute
control of where the tunnel is set up across the
network. GRE, being at the mercy of IP routing,
didn't give them what they really wanted. The
only way to do it with IP routing would have been
source-routing which is more disliked than GRE.

---------------

Actually, GRE has a source routing mechanism that we specified, but never actually implemented. Again, folks told us that the overhead was too high, so we dropped it and moved on. More trivia from the history of routing...

Tony
rtg_dude 12/4/2012 | 9:24:44 PM
re: MPLS Vendors Demo Fast Reroute AAL5,

>>You don't need MPLS to tunnel a packer or a frame
>>across an IP cloud. GRE or L2TP as you suggested
>>can do it really well with no additional state
>>or new sw in the core.
>
> This is really obvious I donG«÷t know why you are
> trying to compare GRE/L2TP to MPLS L3VPNs.

You're missing the point. Your argument was that
MPLS allows SPs to provide L3VPNs. I'm saying
they don't need MPLS for L3VPNs. One can even
have 2547 without MPLS.

> MPLS L3VPNs provide more that tunneling; they
> provides virtual private forwarding table services
> that can be used to create VPNs that forward
> information based on L3 information. I donG«÷t see
> what your argument is here, they are NOT the same.

VRs and VRFs are completely orthogonal to the
tunneling techniques. Your logic is L3VPNs ==
2547 == MPLS. I'm trying to tell you that MPLS was
a design/marketing choice for the tunneling technique
in 2547, but there's nothing in VPN technologies
that calls for it.

> It is technically possible to provide L3VPNs based
> on IP tunneling protocols. IG«÷m not religious about
> MPLS, just being realistic; L3VPNs based on MPLS are
> deployed in customer networks already by Cisco/Juniper,
> I do not see L3VPNs based on IP succeeding.

Thought it works both ways, what vendors do finally
depends on what SPs ask for. I do hear some SPs saying
they need an non-MPLS non-BGP VPN solution. I have
to admit they are in minority. I do hope, however,
that the rest of them do not believe VPNs == MPLS.

>>>>> - Allows SPs to provide L2VPN services to
>>>>why not GRE?
>>> You really mean L2TPv3 don't you?
>>L2TP will work as well.
> Yes L2TP is equivalent but not GRE as you originally said.

You're probably mixing signaling and tunneling together.
One could perfectly have L2VPNs with GRE.
We're diverging from the original point though,
which is again: you don't need MPLS for L2VPNs.

>>Sounds like SPs actually do not need MPLS
>>to provide L2VPNs, huh?...
>
> If a customer has an IP core they can implement point
> to point L2VPNs, but

Let's put a period here. It's important to keep MUSTs
and MAYs separate.

> they can provide extra services
> to customers such as FRR if they implement their L2VPNs
> based on MPLS. Extra services means extra money, if
> SPs find a way of making more money from their customers
> they will choose that technology.
>
> And the IP L2VPN equivalent of FRR isG«™?

VPNs and FRR are orthogonal. You could have an IP-based
VPN and provide MPLS FRR somewhere in your network, or
you could have SONET protection doing this for you.

I wouldn't argue that MPLS FRR is not valuable. I do
think, however, that we should have FRR or fast convergence
in regular IP routing. This is doable and will help
to avoid MPLS complexity.

The benefit of IP-based VPNs is clear: the SPs do not
need to spend money and time on MPLS to provide service,
and the VPN sites or tunnel end-points can be anywhere
in Internet... And the equivalent of the Internet end-to-
end connectivity in MPLS is...?
sgan201 12/4/2012 | 9:24:44 PM
re: MPLS Vendors Demo Fast Reroute http://a1840.g.akamai.net/7/18...

Check this out..
This box support virtual routers.
The virtual router support RIP/OSPF and etc...
MPLS is in there also..
sgan201 12/4/2012 | 9:24:45 PM
re: MPLS Vendors Demo Fast Reroute Is the RPM inside the MGX box??
Yes..

Is this a single box solution??
Yes..
sgan201 12/4/2012 | 9:24:45 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
I think you are a smart person..
Profit = Revenue minus costs..

1) Is people willing to pay more for IP VPN compare to FR??
Please noted that SLA for IP VPN is worst than FR in term of jitter and uptime..
Please noted that according to Infornetics, 60% of the IP VPN is DIY (do-it-yourself).

2) If you design a network, does it costs more or less to provide FR service versus IP VPN services?

3) Please noted that regardless of selling IP or FR over T1, you leased the same T1 from ILEC at same costs..

T1 access to Internet is about $200 to $500 per month..
(Check your daily newspaper)
FR T1 average revenue is $900 per months
(Check Vertical system: 12 billions revenue divided by 1.2 million FR T1 line) ~ 10K per month..

Which one you can charge more and cost less to provide and get you better margin??

You can do this with a simple calculator...

Do your own calculation and derive your own number?? Call a few enterprise network engineer and ask them will they buy IP VPN services..

Just because a SP hyping something, it does not mean it will sell. It is the enterprise network engineers that determine what will succeed or failed.
Remember ISDN, FR SVC, Centrex services and so on..
Enterprise network engineer had the final vote on any services. They are the market..

I was part of the market that decided who live or die in term of services. My division's annual telecom budget was 100 millions. The corporate telecom budget run into billion. Customer is never wrong..
fiber_r_us 12/4/2012 | 9:24:46 PM
re: MPLS Vendors Demo Fast Reroute The Nortel link does not mention anything about the passport being an IP router (i.e. it must forward frames by looking at IP addresses). Without this, it couldn't possibly be an Internet router.
fiber_r_us 12/4/2012 | 9:24:46 PM
re: MPLS Vendors Demo Fast Reroute The RPM is a 7200 on a card with about 300K pps max throughput that supports a FastEthernet and FDDI interface and an ATM port out the backplane... This hardly qualifies even as an access router. This would be the equivalent of using separate 7200s as edge routers with an ATM switch between them.
sgan201 12/4/2012 | 9:24:46 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
You are wrong about Nortel Passport also..
http://www.nortelnetworks.com/...
sgan201 12/4/2012 | 9:24:47 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_Us,
You are wrong.
http://www.cisco.com/en/US/pro...

RPM is the routing module for Cisco MGX..

Goto that URL or search for RPM and MGX
sgan201 12/4/2012 | 9:24:47 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
ATM Switch has the real revenue number but the router folks has the pretend number. But, eventually the real number will come out..
fiber_r_us 12/4/2012 | 9:24:47 PM
re: MPLS Vendors Demo Fast Reroute >ATM Switch has the real revenue number but
>the router folks has the pretend number...

Sorry, I don't quite get this; the "real" and "pretend" number? Are you saying that the service providers that are offering IP services are somehow mis-stating thier IP revenue or ATM revenue?

This gets back to my original question:

If IP/MPLS (the Internet, L2/L3 VPNs, etc) are not profitable services, why are all of the service providers continuing to offer them? Why are the SPs continuing to upgrade the networks? In this economic climate SPs would shutdown an inherently non-profitable service in a heartbeat!

As I stated before, certain self-serving groups in SPs would like everyone to believe that IP/MPLS services are not profitable just so they can continue to push the only technologies they know and understand (ATM, FR, TDM, voice, etc). I have worked in several carriers (the largest ones) and have seen this "mud slinging" first hand. Instead of trying to improve existing services, they attack any sort of new service with FUD tactics with no facts to back them up. This is where the "Internet is not profitable" came from.
fiber_r_us 12/4/2012 | 9:24:48 PM
re: MPLS Vendors Demo Fast Reroute Dreamer said:

>Actually, I am not very inclined/interested
>in convincing anyone.

Ahh... that would explain your high volume of posts promoting the everything over ATM approach. If you are not trying to convince anyone else, then you must be trying to convince yourself.

>When FR/ATM service revenue exploded and ATM
>Switch revenue overtaking core and edge router revenue...

Umm, I thought you already said that the IP router revenue is nowhere close to ATM/FR revenue?
fiber_r_us 12/4/2012 | 9:24:49 PM
re: MPLS Vendors Demo Fast Reroute Dreamer,

The MGX series is a FR/ATM switch. It does not support IP routing (forwarding based on IP addresses, OSPF, ISIS, BGP, Multicasting, etc) and thus could never support the Internet. It also apparently does not offer support for POS or Ethernet interfaces. Its MPLS support is limited (at best) to ATM and PNNI in cell-switching mode. This does not qualify as a platform that supports all of these services previously stated.

Similarly, the Nortel Passport 15K/20K is purely a Layer 2 device with no support for layer 3 packet routing.

The Alcatel 7670 info on the web page makes it appear as this device may qualify as a device that offers all of the stated services. But, I don't know how much of the functionality mentioned on their web page is future and how much is here now.
sgan201 12/4/2012 | 9:24:52 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
Cisco MGX..
It is not a core router but it does not have to be..

Nortel Passport 15K/20K

Alcatel 7670 RSP is an ATM switch
sgan201 12/4/2012 | 9:24:52 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
Actually, I am not very inclined/interested in convincing anyone. I am laughing my way to the bank. The more people waste their time and money on MPLS, IP VPN, and so on. The better off I am..

"The truth will set you free"

Eventually, the revenue figure will come out and the market will decide who will survive and who will not. All the SP and equipment manufacturer that is not smart enough will die and they are dying now..

Being an ATM bigot that had been hammerred by those IP/MPLS folks, this is payback time..

When FR/ATM service revenue exploded and ATM Switch revenue overtaking core and edge router revenue, the truth will come out..
sgan201 12/4/2012 | 9:24:53 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
Your statement about ATM Switch does not support IP VPN service and IP routing is not true..
Alcatel, NT, Cisco all offred embedded router module that support that..
If your argument about MPLS is cheaper because of single box/single core solution than both ATM switch with embedded router module that support MPLS and MPLS switch is even..
The only difference is ATM Switch can support DS1/E1 TDM service now...
sgan201 12/4/2012 | 9:24:53 PM
re: MPLS Vendors Demo Fast Reroute Hi cc_junk,
One CE = one customer locations..
Assuming all those numbers are IP VPN only, you still come up a very small number..
Ditto on Savvis, one VR = one customer (for many locations)..
Equant and Savvis is not the top 3 but they have about 10%+ of the market..
If 10% = X,
100% = 10X, you can get an idea how many locations/customers..
In your study, what is market share for Equant and Savvis??
fiber_r_us 12/4/2012 | 9:24:53 PM
re: MPLS Vendors Demo Fast Reroute Dreamer,

I am unfamiliar with Cisco's offering of an ATM switch that routes IP, supports MPLS at scales that would be useful to support a large carrier's Internet core. Is this some new form of GSR? Which product is this?

NT offers an IP/MPLS router suitable for a large Internet backbone???

Alcatel is trying with the 7670... But this is not an ATM switch.

fiber_r_us 12/4/2012 | 9:24:54 PM
re: MPLS Vendors Demo Fast Reroute cc_junk:

Dreamer uses the Bobby Max market analysis approach! :-)

Sorry dreamer, couldn't resist. But, you should be more explicit on how you came up with your numbers.
cc_junk 12/4/2012 | 9:24:54 PM
re: MPLS Vendors Demo Fast Reroute dreamer101:

Where are you getting your market info?

The link you provided
http://ppvpn.francetelecom.com...

Has a presentation by Equant discussing the number of CEs, and since they don't clarify it looks like they are counting both normal FR customers and L3 VPN.

The Savvis presentation says little other than mentioning the port numbers for their one big customer (Rueters).

No survey there. No info about BT, AT&T, Sprint, Worldcom.

The enterprise surveys by the standard consulting agencies I have seen show Worldcom and AT&T as the top VPN providers and those count both Internet tunneling VPNs and networks based L3 VPN.

AAL5 12/4/2012 | 9:24:55 PM
re: MPLS Vendors Demo Fast Reroute >>> can you tell us what value MPLS adds for L2
>>> and L3 VPNs compared to what IP already does?
>
>> L2 point to point tunneling based on MPLS i.e.
>> Martini compared to L2TPv3 has similar functionality.
>> What IP based protocol has similar functionality to
>> MPLS L3VPNs so I can do a comparison?

>You don't need MPLS to tunnel a packer or a frame
>across an IP cloud. GRE or L2TP as you suggested
>can do it really well with no additional state
>or new sw in the core.

This is really obvious I donG«÷t know why you are trying to compare GRE/L2TP to MPLS L3VPNs.
MPLS L3VPNs provide more that tunneling; they provides virtual private forwarding table services that can be used to create VPNs that forward information based on L3 information. I donG«÷t see what your argument is here, they are NOT the same.


>>> can you point out those SPs who have no BGP in
><> the core
>
>> No I can't, customer's internal network configuration
>> information is confidential.

>Let's make it simpler: how many of them do you know?

2 Major SPs.

>>>> - Allows SPs to provide L3VPN services
>>>> (revenue generating)
>
>>>why not GRE?
>
>> You're kidding right? They do not provide similar
>> services, please look up a tutorial on MPLS L3VPNs
>> to get an idea of the difference between these and
>> GRE tunneling.

>Remember you said that MPLS allows SPs to provide X
>Read 2547 and notice the difference between control
>plane and tunneling. The story is (not that I
>believe it) that control plane mechanisms (BGP) are
>transparent to the core; the tunneling part, on
>the other hand, requires MPLS. Now, L3VPNs do not
>need MPLS for tunneling. One could do L3VPNs with
>the BGP hack from 2547 (not to say that I like it much)
>and GRE, in fact I think I've heard RB does this
>already.

It is technically possible to provide L3VPNs based on IP tunneling protocols. IG«÷m not religious about MPLS, just being realistic; L3VPNs based on MPLS are deployed in customer networks already by Cisco/Juniper, I do not see L3VPNs based on IP succeeding.


>>>> - Allows SPs to provide L2VPN services to
>>>> customers based on their
>>>> existing IP network (revenue generating and
>>>>reduces capex)
>
>>>why not GRE?
>
>> You really mean L2TPv3 don't you?

>L2TP will work as well.

Yes L2TP is equivalent but not GRE as you originally said.

>> There is no advantage, for
>> providers that run a IP only core L2TPv3 is ideal
>> for them to provide L2VPN services.

>Sounds like SPs actually do not need MPLS
>to provide L2VPNs, huh?...

If a customer has an IP core they can implement point to point L2VPNs, but they can provide extra services to customers such as FRR if they implement their L2VPNs based on MPLS. Extra services means extra money, if SPs find a way of making more money from their customers they will choose that technology.

And the IP L2VPN equivalent of FRR isG«™?

AAL5
fiber_r_us 12/4/2012 | 9:24:56 PM
re: MPLS Vendors Demo Fast Reroute Dreamer,

My response was somewhat "tongue-in-cheek" as *you* offered no proof of why ATM/FR was cheaper than MPLS in your previous message (other than MPLS was layer "3" and ATM/FR was layer 2). This is really not proof. I fully anticipated your "prove it" response. The "prove it" mantra will not get us anywhere as neither side can "prove" anything to the satisfaction of the other. As all of us know "your mileage may vary" is the phrase that one must keep in mind when comparing network architectures. There are so many variables in any real analysis that it is pointless to try to "prove" it on a discussion board with made-up network architectures. The only comparison that really matters is in a given provider's real network, with real equipment, real services, real people, and only at a given point in time. When using any other approach, it is like statistics, you can "prove" whatever you want.

The "proof" I offered was the fact that you could run all of the Layer 2 services (ATM, FR, Ethernet) over an MPLS network (of course ATM could do this), *AND* you could run your Internet and Layer 3 VPN traffic as well (which ATM could not do by itself). Hence, one infrastructure would provide for a more cost-effective network both capitally and operationally.
sgan201 12/4/2012 | 9:24:57 PM
re: MPLS Vendors Demo Fast Reroute Hi Skeptic,
I would suggest you read ATM Forum standard on doing CES over ATM. It is not a walk on a park.. Clock synchronization is a major problem.. To do the same thing on MPLS without going into cell mode is probably possible if all your uplink is OC-48 or higher. If that is what MPLS switch is designed for, then ATM folks will laugh our way to the bank.. If MPLS goes into cell mode, then what is point of doing MPLS when ATM can hanlde that cheaper..
There are 100,000 CO in USA. Most of them are connected via OC3/OC-12.
sgan201 12/4/2012 | 9:24:57 PM
re: MPLS Vendors Demo Fast Reroute Hi Sketic,
I used to be Enterprise Network Engineer for a fortune 50 company. Believe me when I said we do pass real-time transaction between business locations via FR. In fact, that is the main reason to inter-connect business location. If not, we could just buy internet access for every location and do-it-yourself (DIY) IP VPN..
sgan201 12/4/2012 | 9:24:58 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
1) It costs more for the SP to provide FR/ATM than MPLS..
Where is the proof??
All the services that you mention can be provided by ATM core.

Plus, MPLS cannot support DS1/T1 leased line which ATM Core can.

2) Core is only 40% of the costs.. Access is the 60% of the costs.

3) Worldcom launch the bundled FR services. Everyone on the market has to matched the price including IP VPN.

4) My point about Worldcom bundled FR service is that it is provided with very low access costs to worldcom via FR NNI interface. This option is not available on current MPLS switch..
skeptic 12/4/2012 | 9:25:00 PM
re: MPLS Vendors Demo Fast Reroute A historical note: GRE certainly preceeded MPLS, but was roundly ignored by the ISPs because of its overhead.
-------------------
From what I remember, it was ignored because
how certain vendors implemented it in their
forwarding. That it was all done on the slow
path meaning that GRE could not be used as more
than a toy. That may be fixed in newer equipment
(I dont know), but the perception is still there.

Also, most people who do tunneling want absolute
control of where the tunnel is set up across the
network. GRE, being at the mercy of IP routing,
didn't give them what they really wanted. The
only way to do it with IP routing would have been
source-routing which is more disliked than GRE.
skeptic 12/4/2012 | 9:25:00 PM
re: MPLS Vendors Demo Fast Reroute Are you saying IP routing = IP Service??
To me, Internet = Basic IP routing is not equal to IP service..
What is a profitable IP service that a service provider can sell??
-----------------------
You made the arguement that IP is a "religon"
not based on being a "real business". If IP
is a religion, then it would follow that there
is no need for it. But the existance of the
internet suggests that there is a real need
for it. So rather than narrowly focusing on
"profits for services", your going to have to
explain:

a) Why IP and the internet, being a religion,
are not necessary.

b) How basic IP service would be delivered
without an IP backbone and IP routers.



=======================
Internet is the best thing that IP can do.,.
But, it is best effort and it is not suitable for any reasonable mission critical business stuff.
By trying to stretch IP via MPLS to deliver good QOS which it is never intended to do, IP had failed..
-------------------------------------
Yes. But if you look at whats on private-line
service today, the majority of it isn't mission
critical real-time QOS sensitive traffic.

Transaction processing is obviously a R/T
application. But connecting several corparate
sites isn't.

And there is no technical barrier to doing ATM
style private line services with MPLS. You
need a highly available router, you need to
keep the MPLS private-line traffic seperate
(different links) from the IP traffic and you
need a few features for QOS enforcement in the
forwarding path. The great thing about MPLS
is that you can create a dynamic tunnel through
the network that only depends on IP for signaling.






fiber_r_us 12/4/2012 | 9:25:02 PM
re: MPLS Vendors Demo Fast Reroute Hi Dreamer101:

1) It costs more for the SP to provide FR/ATM service than MPLS. Since you can combine all "data" services (FR, ATM, Ethernet, L3 VPNs, Internet, packet voice, packet video, etc.) into MPLS and maintain one infrastructure. This allows you to maximally utilize your hardware infrastructure while minimizing your operational costs (management systems, power, space, employees, spares, etc).

2) FR/ATM VPN does not provide added value over MPLS, as QOS can be achieved either way.

3) Any service can be "managed".

4) Using Worldcom and an example of providing profitable services is damaging your credibility. But, even Worldcom is migrating to MPLS.
rtg_dude 12/4/2012 | 9:25:03 PM
re: MPLS Vendors Demo Fast Reroute Tony

> ... If you add 28 bytes, then you're at about 40%
> overhead. Most people found that objectionable and
> petitioned for a way to do tunneling (and source
> routing) with lower overhead.
>
> The only way to do that was to go to a connection
> oriented approach, and therefore MPLS.

-and that boils down to the dilemma of control
plane simplicity/scalability vs bandwidth
efficiency once again.

SPs should to compare the costs of MPLS
deployment and operations to the cost of the
bandwidth they need to provision. This is true
for both VPN encapsulation and TE parts of the
story.

Note, by the way, that SP would have no control
over VPN encapsulation overhead in the case of
CE-based VPNs (not to say that they shouldn't
care when they actually can control it).

Another point that VPN folks should keep in mind
is the fact that there is only one way to carry
data (including a VPN packet) from *any* place on
the Internet to *any* other place on the
Internet - send an IP packet. Attempts to stretch
MPLS-based VPN schemes to multi-SP scenarios will
inevitably result in IP-over-MPLS-over-IP-over-L2
kludges, that, by the way, make you look at the
bandwidth argument from a different angle.
rjmcmahon 12/4/2012 | 9:25:04 PM
re: MPLS Vendors Demo Fast Reroute IP services is a misnomer. The telco world first started misusing the term "services" many years ago as they now misuse the term broadband. Redefining terms is common trick used by those attempting to deceive people into believing things which aren't true.

IP provides a network for networks. Such a digital networks primary purpose is to interconnect networks so bits can move freely. Moving bits is a transport and distribution business.

Such revenues will come from transporting bits of value and delivering bits in volume -- not much different than a transmission and distribution network used for our electricty, a multimodal container network used for our physical goods, or a post system used to deliver script on paper.
sgan201 12/4/2012 | 9:25:05 PM
re: MPLS Vendors Demo Fast Reroute Hi cc_junk,
I am sorry to tell you that for Savvis, even with that kind of revenue, it is one of the largest IP VPN provider with about (10%+) of the market share..
In fact, in term of number of IP VPN customers, number of locations, this network is comparable to Equant's networks..
Now, you can see how big a joke this is..
sgan201 12/4/2012 | 9:25:05 PM
re: MPLS Vendors Demo Fast Reroute Hi cc_junk,
http://ppvpn.francetelecom.com...

At this URL, you can see what is the number of customer and locations for each providers..

Savvis is quite big..
sgan201 12/4/2012 | 9:25:06 PM
re: MPLS Vendors Demo Fast Reroute Hi cc_junk,

1) It cost more for the SP to provide IP VPN service than FR. (Laye 3 1/2 versus layer 2)

2) IP VPN does not provide value add as compared to FR (in fact it is value minus since it is lower QOS).

3) Managed service can be provided via either FR or IP VPN..

4)Worldcom use of FR NNI interface can sell FR at very low costs and still make money.. The reason is that instead of buying 2 T1 line from ILEC, Worldcom buy intra-LATA FR DLCIs from ILEC to inter-connect into Worldcom inter-LATA network. This lower their costs substantially..

5) With (4), FR is at parity or lower costs than IP VPN in full meshed configuraion: time for IP VPN to die..
sgan201 12/4/2012 | 9:25:06 PM
re: MPLS Vendors Demo Fast Reroute Hi lob,
Do you drink water from the tap or you buy bottled water??

Is it acceptable for you to go an ATM machine to withdraw your money and sometimes it take 20 to 30 seconds to complete your tranaction and sometines 2 seconds. Or, it always has to be 2 seconds..

Answer the above two questions and you will have the answer..

Why should I sacrify QOS when it is the same price or lower than without QOS??
Tony Li 12/4/2012 | 9:25:06 PM
re: MPLS Vendors Demo Fast Reroute
A historical note: GRE certainly preceeded MPLS, but was roundly ignored by the ISPs because of its overhead.

If you look at IP traffic, 40% of the packets are 40 bytes (TCP acks). If you add 28 bytes, then you're at about 40% overhead. Most people found that objectionable and petitioned for a way to do tunneling (and source routing) with lower overhead.

The only way to do that was to go to a connection oriented approach, and therefore MPLS.

Tony
cc_junk 12/4/2012 | 9:25:08 PM
re: MPLS Vendors Demo Fast Reroute lob in post #76:

"(There, of course, is the argument stating that circuit routing is more flexible since it can route around already filled pipes. In fact, this is not important, since most real backbones are designed to match topology to traffic patterns)."

Taking a tangent from the thread using the above quote.

I have seen very large ATM networks for FR/ATM services designed that way: matching topology to traffic patterns. Such networks have many direct trunks and high meshedness resulting in high node degrees and fewer hops. They can get away with this because they have constrained routing. Whereas very large IP networks require very regular designs with heirarchical hubbing. The topology does not match the traffic matrix as closely. They end up with smaller node degrees, higher bandwidth trunks, more hops, larger switching fabrics. So, putting aside the availability question, can someone answer whether a network was designed with MPLS TE along the lines of ATM designs could be significantly less costly then the non-TE network for the same traffic matrix? Better, assume that the traffic matrix is growing and evolving so that it may not always directly map onto an initial hierarchical hubbing design.

If the carrier had its own optical switching STS1 infrastructure with channelized interfaces on the routers, it would mitigate the cost difference between say 4 discrete OC12s and a single OC48. Then the tradeoff will the more expensive larger fabric+interface nodes versus less expensive smaller fabric+interface nodes. Will this matter in the cost equation?

rtg_dude 12/4/2012 | 9:25:08 PM
re: MPLS Vendors Demo Fast Reroute >> can you tell us what value MPLS adds for L2
>> and L3 VPNs compared to what IP already does?
>
> L2 point to point tunneling based on MPLS i.e.
> Martini compared to L2TPv3 has similar functionality.
> What IP based protocol has similar functionality to
> MPLS L3VPNs so I can do a comparison?

You don't need MPLS to tunnel a packer or a frame
across an IP cloud. GRE or L2TP as you suggested
can do it really well with no additional state
or new sw in the core.

>> can you point out those SPs who have no BGP in
>> the core
>
> No I can't, customer's internal network configuration
> information is confidential.

Let's make it simpler: how many of them do you know?

>>> - Allows SPs to provide L3VPN services
>>> (revenue generating)
>
>>why not GRE?
>
> You're kidding right? They do not provide similar
> services, please look up a tutorial on MPLS L3VPNs
> to get an idea of the difference between these and
> GRE tunneling.

Remember you said that MPLS allows SPs to provide X
Read 2547 and notice the difference between control
plane and tunneling. The story is (not that I
believe it) that control plane mechanisms (BGP) are
transparent to the core; the tunneling part, on
the other hand, requires MPLS. Now, L3VPNs do not
need MPLS for tunneling. One could do L3VPNs with
the BGP hack from 2547 (not to say that I like it much)
and GRE, in fact I think I've heard RB does this
already.


>>> - Allows SPs to provide L2VPN services to
>>> customers based on their
>>> existing IP network (revenue generating and
>>>reduces capex)
>
>>why not GRE?
>
> You really mean L2TPv3 don't you?

L2TP will work as well.

> There is no advantage, for
> providers that run a IP only core L2TPv3 is ideal
> for them to provide L2VPN services.

Sounds like SPs actually do not need MPLS
to provide L2VPNs, huh?...

cc_junk 12/4/2012 | 9:25:08 PM
re: MPLS Vendors Demo Fast Reroute

"..3) Virtual private networking services in the form of Ethernet, FR, and ATM VPNs via Martini over an MPLS core. These are clearly a "data service" of some form; more like a "MPLS service". Whether you wish to call them an "IP service" is a matter of semantics."

These services are prettry much indistinguishable from legacy FR/ATM/TLS services implemented over ATM, at least from the service interface the user sees. I certainly would not call them IP services. The interface definition to the user is layer 2 UNI and is not IP specific. The service interface can carry IPX, SNA, packetized voice, among others, including IP.

"Unfortunately, all of the services combined have yet to generate enough income to easily satify the "I" (Interest) part of the EBIDTA equation"

Interesting view. The "I" may overwhelm. Did you glean that they are operationally profitable form the quarterly reports?

I don't trust providers are playing fair with their operational costs, certainly not after Worldcom. For example, if a carrier decides to deploy a new or upgraded platform they might try to capitalize the migration costs of moving service. If they insert the new equipment between the old and the user (in the middle of the customer access) and then later remove the old they may try to view the evolution as capital. If they deploy the new separately and then initiate migration moves (reterminating the access circuit) from the old to the new they will have to cost it as operations. The network will look the same in the end but they will cost a piece of it (user migration) different ways.


cc_junk 12/4/2012 | 9:25:09 PM
re: MPLS Vendors Demo Fast Reroute [I hope this message wordwraps. Got tired of trying to manually wrap.]

dreamer101 in #58:
"The most popular network based L3 VPN service is IP enabled FR or IP over ATM with VR. Look at Savvis implementation: IP over ATM with VR.

I didn't think Savvis was a major enterprise WAN carrier. Aren't they a specialized vertical provider for the financial industries? They are a company whose 2nd quarter revenue outside of their Rueters business was $16 million and of that the managed IP VPN sservice was just $4 million and shrunk from Q1 to Q2. Not a implementor of L3 VPN with the ability to fundamentally change carrier enterprise WAN services.

As for the other L3 VPN services, all the major ones (Worldcom Private IP, C&W, AT&T, Equant/FT, BT) in the market are 2547 based. Sprint, late to the game just some months ago, is initially doing VR with PEs interconnected with IPSEC (?), yet they are supposedly developing 2547 for deployment.

Anyone know of RBOC plans for network based L3 VPNs? Have seen some info that Bell South is developing 2547 VPN service.

Why are these services a "joke" if they are growing a huge rates? Just because they aren't even a tenth of the FR market yet, doesn't make them a joke. There are tens of thousands of ports in service. They are growing at more than 100%. These services aren't much more than two years old and they are probably where it took FR to reach in four or more years (maybe FR in 1994 or 95).

And it is not important whether they "surpass FR". If they can generate positive margins (who knows yet?) and their revenue grows to a significant share compared to FR (25% ?) then I think anyone would consider it be a success? (wouldn't you consider a multi-billion dollar market hopefully with positive margins a success?) Of course it will take a few years, but only a dot-commer from 1999 would think you could create a new multi-billion dollar business in a year or two. For a carrier already offering FR they want to position these services to winback enterprises from their competitors. I doubt they will migrate their existing customers if they can help it. But they will if it is what it takes to retain the customer. And if the margins aren't positive, the vendor watchers in this forum could care less as long as there is a lot of equipment/upgrades to be sold.

Now as for pricing. You bring up a very good point that FR services can lower their prices as Worldcom has done (do you believe that Worldcom has positive margins at those prices? I hope Worldcom doesn't try to drive FR into a commodity space). As I said in my previous post, I do not see enterprises switching to L3 VPN for price. Many providers in fact choose to price it slightly higher than FR because they feel it brings other value adds over FR. Enterprises that want those value adds may switch without the direct price benefit (they may see cost benefits from other indirect effects like being able to eliminate their legacy FR aggregation layer in front of their hubs). Actually, I anticipate that the big FR providers will lower their prices to counter a large movement of enterprises to do-it-yourself Internet tunneling VPNs. If they start to see wholesale technology replacement then they will just lower their prices. They will also lower their L3 VPN service prices. Indeed why do you think the carriers are rushing so quickly to embrace network-based VPNs? For Internet tunneling VPN the provider only sells a commodity Internet access. This is not a space the carriers want to end up in (see Business Communications Review, Feb 02). They only sell enterprises Internet VPN when the provider gets the value add of managing the enterprise WAN.

"IP VPN does not stand a chance in USA." A chance at being successful? Well, it does. And I would say that it is well on the way to success. Let's see if the providers can do it profitably at scale.

AAL5 12/4/2012 | 9:25:09 PM
re: MPLS Vendors Demo Fast Reroute rtg_dude:

> can you tell us what value MPLS adds for L2
> and L3 VPNs compared to what IP already does?

L2 point to point tunneling based on MPLS i.e. Martini compared to L2TPv3 has similar functionality. What IP based protocol has similar functionality to MPLS L3VPNs so I can do a comparison?

> can you point out those SPs who have no BGP in
> the core

No I can't, customer's internal network configuration information is confidential.

>> - Allows SPs to provide L3VPN services
>> (revenue generating)

>why not GRE?

You're kidding right? They do not provide similar services, please look up a tutorial on MPLS L3VPNs to get an idea of the difference between these and GRE tunneling.

>> - Allows SPs to provide L2VPN services to
>> customers based on their
>> existing IP network (revenue generating and
>>reduces capex)

>why not GRE?

You really mean L2TPv3 don't you? There is no advantage, for providers that run a IP only core L2TPv3 is ideal for them to provide L2VPN services.


AAL5
cc_junk 12/4/2012 | 9:25:10 PM
re: MPLS Vendors Demo Fast Reroute techstud posting 62:
"I am surprised how intelligent people like you on this list will
come to wrong conclusions. So called, BGP/MPLS VPN model has very
little to do with mpls ..." and then goes on to talk about how
the 2547 model could be implemented using other forms of tunneling
between PEs.

You completely misread my message (posting 57). I know as well
as anyone how 2547 is implemented. Did you not read the entire
message? You didn't quote the next paragraph after the one you
did quote:

from message 57
"L3 VPN's are being implemented using MPLS, although it need
not be. Just as FR is implemented using ATM. Nevertheless,
it is the implementation that has gained mind/market share,
so any alternative implementation won't get off the ground
because it will offer no sufficient benefit to switch midstream
this late in the game. Even though the L3 VPN user does not
see MPLS directly, the users and market view this as the MPLS
service/network."

As you see, I acknowledged that MPLS is not visible to the user,
say that it need not be implemented using MPLS and that other
implementations (without enumerating the obvious) were possible.
I said that the MPLS implementation had the mindshare and market.
It is probably too late for IP in IP, IPSEC or GRE between PE to PE
to overcome the commitments already made.

C&W, AT&T, BT, Equant, Worldcom, DT (not certain about availability)
are all selling 2547-based VPNs using MPLS (the Cisco implementation),
plus many small time carriers. Sprint has a VR based service (other vendor,
Cosine?) with IPSEC between the PE VRs (not really scalable for a full
mesh peering among VRs in a VPN because Sprint is not creating VRs per VPN
in the P devices). And I understand that Sprint is developing the 2547
version for deployment.

In fact it is interesting that the major VR vendors (Nortel, Redback,
Cosine) are now marketing 2547 implemented using VRs instead of vrfs.
If VRs with tunnels were what the carrier market was implementing these
guys would not have had to modify/enhance their portfolios.

There are many things I like about the IP in IP tunneling approach between
PE's for 2547. You avoid having to deploy MPLS and its control plane.
You can use standard area designs and summarization (with MPLS you must
have a host route to every other PE in the AS). But as far as I know no
vendor has implemented it in a form that is automatic (as described in
draft-ietf-ppvpn-gre-ip-2547-00.txt section 6.1) like the MPLS implementation.
And until they do, configuration management is actually more complex
than with using MPLS. Specifically, deployable implementions using other types
tunneling require you to create an explicit tunnel interface for every other
PE in the network - makes incremental additions of PE to the network error
prone. Also, you lose the TE capability of MPLS (personally I don't think
the TE is necessary - perhaps in a given carriers international footprint
where they have more capacity constraint, but legacies could probably use
ATM in those areas).

But as I said previously, the alternative implementations likely will lose
since the 2547 with MPLS implementation is already way down the road.

fiber_r_us 12/4/2012 | 9:25:10 PM
re: MPLS Vendors Demo Fast Reroute
>A) Wholesale Dial-up..

The vast majority of this (90+%) is dial-up Internet access running over an MPLS backbone. Sounds pretty IP'ish to me.


>B) Wholesale bandwidth (TDM)

Along with wavelengths and fiber, private line (TDM) are traditional services that I would not qualify as "IP services" or "data services"; these are more like "transport services". However, the vast majority of the traffic flowing over these transport services is "data" in the form of IP packets.

>C) Wholesale bandwidth IP
>D) Wholesale bandwidth ATM/FR

There are several things here:

..1) Internet Access in the form of DIA (Dedicated Internet Access) where DS-3, OC3, OC12, and OC48 ports are sold as access to Internet routers running over an MPLS core. Clearly this is an "IP" service.

..2) Internet Access in the form of Ethernet pipes in Colocation spaces where 10, 100, and Gigbit Ethernet ports are sold as access to Internet routers runinng over an MPLS core. Clearly another "IP" service.

..3) Virtual private networking services in the form of Ethernet, FR, and ATM VPNs via Martini over an MPLS core. These are clearly a "data service" of some form; more like a "MPLS service". Whether you wish to call them an "IP service" is a matter of semantics.

Level 3 also sells colocation, which would be a service in itself.

>From what I know, only A is highly profitable
>and contribute majority of the revenue.

All are profitable (in the EBIDTA sense), with none being more than 50% of the revenue stream.

Unfortunately, all of the services combined have yet to generate enough income to easily satify the "I" (Interest) part of the EBIDTA equation (though they seem to be on the right track).

The "wholesale" phrase that you use really only means that they deal with large scale customers. This includes other providers (which is where the wholesale comes in), but also large scale corporate and government customers. They just don't deal with individual end-users.
lob 12/4/2012 | 9:25:11 PM
re: MPLS Vendors Demo Fast Reroute I've seen someone claiming that "the best effort" delivery is insufficient for "critical business apps" and some form of resource reservation is needed.

This is a very common misconception. The fact is, "the best effort" is _more_ reliable than resrved bandwidth schemes, which is easy to see if you follow this proof:

1) if network is operating _under_ its maximal load, there's no difference between reserved bandwidth and pure best effort, all packets get through.

2) if there's insufficient capacity, the best effort (particularly when FQ/RED is used) spreads the degradation fairly between all sessions, thus ensuring continuing universal service availability.

This contrasts with per-virtual circuit bandwidth reservation which would fail to allocate bandwidth for some circuits - thus _denying_ some service completely.

QoS on its own does not create any extra bandwidth by magic; all it does is modifies the notion of fairness.

There is no such thing as "critical business" which would benefit from absense of communication whatsoever instead of degraded communication.

The correct question is not how to allocate bandwidth, but how to provide premium service: this is very easily achieved by using IP priority bits to select between fixed queueing/drop priorities, and by policing inputs to prevent customers from abusing high-priority priveleges.

(There, of course, is the argument stating that circuit routing is more flexible since it can route around already filled pipes. In fact, this is not important, since most real backbones are designed to match topology to traffic patterns).
sgan201 12/4/2012 | 9:25:11 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
What IP Services do Level 3 sales??

A) Wholesale Dial-up..

B) Wholesale bandwidth (TDM)

C) Wholesale bandwidth IP

D) Wholesale bandwidth ATM/FR

In your definition, which one is IP service?
Which one is data services??

Fromn what I know, only A is highly profitable and contribute majority of the revenue..
DO you call A IP services???
rtg_dude 12/4/2012 | 9:25:12 PM
re: MPLS Vendors Demo Fast Reroute techstud

> ... I don't know why a user of VPN will
> care if the tunnel is MPLS or GRE or IP-in IP or
> whatever as long as it works.

because they go to Networkers ;)

> A VPN will work fine with any tunneling protocol, presence
> of mpls in the network has no special benefits. Doesn't it
> add more compexity and operational cost?

as long as this helps vendors sell their boxes,
do you think they care? as far as SPs go, it takes
a lot of courage to stand up against MPLS given
that you have to fight everyone around.
rtg_dude 12/4/2012 | 9:25:13 PM
re: MPLS Vendors Demo Fast Reroute AAL5-

> If some people think service providers are not asking for MPLS
> enabled switches/routers they are living in a delusional world.

they are, the question is why

> The fact is that major SPs have said that they see their ATM/FR
> networks transitioning into using an MPLS infrastructure for L2 and
> L3 VPNs.

can you tell us what value MPLS adds for L2 and L3 VPNs
compared to what IP already does?

> Using MPLS gives them the following benefits:
>
> - Takes BGP out of the core, reducing routing convergence/routing
> flaps (i.e. enhances routing stability for core data network)

can you point out those SPs who have no BGP in the core, and
while you are at it, could you tell us how they would scale
their iBGP mesh without the internal BGP speakers?

> - Allows SPs to provide L3VPN services (revenue generating)

why not GRE?

> - Allows SPs to provide L2VPN services to customers based on their
> existing IP network (revenue generating and reduces capex)

why not GRE?

> - Allows Opex for FR/ATM/IP networks to be amalgamated into one
> network that provide regular IP L2VN and L3VPN services.

would you mind showing us the other side of the equation where the
cost of such transition and operation of such a converged
network is shown?

fiber_r_us 12/4/2012 | 9:25:13 PM
re: MPLS Vendors Demo Fast Reroute Uhh... I think I already noted that Level3's problem was that thier revenue was insufficient at its current levels to support thier debt load. They accumulated billions in debt during the buildout with the anticipation that thier volume of network sales would be larger than it is now. The cost of the buildout was *not* the IP/MPLS gear, it was the 10,000s of mile of conduit and fiber systems deployed. Even with the massive debt load, it would not require thier data revenue to be dramatically higher than it is now to make thier debt supportable.

In the end, this illustrates that thier data services are highly profitable (70+% gross margin as stated in thier reports). Thus, the IP services themselves *are* profitable and are supporting (almost) the debt load of the rest of the network.
sgan201 12/4/2012 | 9:25:14 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
What is so hard about find out the profitability of a IXC/non-ILEC service??
60% of the costs providing inter-LATA data service is on the ILEC T1 line that you have to leased from ILEC. The whole sale tariff is public information..
Only 40% of the costs is dominated by the backbone network..
Given that regardless whether you sell T1 line as leased line or FR or ATM or IP, your costs is fixed..

A) The average FR revenue per T1 port is 2K per month..
B) The T1 to Internet is only $500 per month..

Given that 60% of the costs is fixed and in A, you can charged 24K per year and in B, you can only charged 6K per year. Isn't it simple arimethic that, A is more profitable than B??

In fact, you can also find out how many T1 is sold by each ILEC to each ISP/IXC..

"The truth will set you free"

Are you dare to face the truth??
sgan201 12/4/2012 | 9:25:14 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
It is not difficult to find out the truth...
The question is do you want see it..
Level 3 revenue as per recent quarter excluding software is only 274 millions.
http://www.lightreading.com/do...

Please noted again Level 3 is only a wholesaler..
It does not sell direct to the customer..
For SP, the 20% CAPEX rule applies..
Given the total revenue for Level 3 is only about one billion dollar a year. Level 3 should only spend 200 million on CAPEX per year or else bankruptcy..
It does not matter how well your network is..
If your violate the 20% rule, your cash flow will kill you..
The business model does not work because there is insufficient IP revenue..
dljvjbsl 12/4/2012 | 9:25:15 PM
re: MPLS Vendors Demo Fast Reroute Then again
what new heavyweight protocols are being
developed that need formal methods? None


SIP, Megaco, SDP

Indeed witht eh rsie of XML. protocol development is nowthe provice of teh masses. There is a gaping hole where protocol verfication technqiues suitable for the ordinary designer should be. Check out the URN question in ITU SG17

dljvjbsl - 1972 - 1979
fiber_r_us 12/4/2012 | 9:25:15 PM
re: MPLS Vendors Demo Fast Reroute I think it would be difficult to provide "hard" numbers illustrating *any* service is profitable or not. The most you have to go on are the publically reported per service numbers. Since the services that providers offer are built on top of multiple technologies, it is extremely difficult to say what is profitable or not at the service level. All per service numbers are thus highly suspect. For example, take the oc-3 that is the "private line" between a customer and an Internet router: is that oc-3 "private line" revenue or "Internet" revenue? Are the expenses for it attributed to "private line" or "Internet"? There is a huge amount of politics inside of the carriers to "prop up" the numbers of traditional services by burdening the next-gen services.

Level 3 has stated that thier revenue for thier data services (IP and MPLS) provide for 70% gross margin. Great gross margin for any product. Unfortunately, thier debt load from building such a large international fiber plant puts a damper on such healthly gross margins unless the volume is large enough.

Well designed MPLS-based backbones deliver as good a QOS as other technologies for a lower capital and lower operating cost. Hence why all of the carriers are migrating that way.

I have no idea why Cosine has so little revenue. Maybe they don't have a differentiated enough product and it was introduced into the worst market in history. Pretty much describes most startups at this point. But what has this got to do with the profitability of service provider networks?
sgan201 12/4/2012 | 9:25:16 PM
re: MPLS Vendors Demo Fast Reroute Hi Fiber_r_us,
It is very simple..
Counter my argument with real figure.
What is the IP service revenue for those SP??
What is the profit margin??
How many customer do they have??

I have a very basic argument why should a enterprice pay for deliver X number of bits between point A and point B with lower QOS??
IP VPN service costs more than FR..
With the new Worldcom bundled FR offer, IP VPN costs more even in the case of full mesh..

"The truth will set you free"

If IP service is really doing well why Cosine has so little revenue??


sgan201 12/4/2012 | 9:25:16 PM
re: MPLS Vendors Demo Fast Reroute Hi Skeptic,
Are you saying IP routing = IP Service??
To me, Internet = Basic IP routing is not equal to IP service..
What is a profitable IP service that a service provider can sell??

Internet is the best thing that IP can do.,.
But, it is best effort and it is not suitable for any reasonable mission critical business stuff.
By trying to stretch IP via MPLS to deliver good QOS which it is never intended to do, IP had failed..

The horrible thing is that there is one and only way that can deliver absolute QOS for IP now. It can be deployed and it works. But, because it is IP over ATM and the IP folks get religios and reject it outright..
skeptic 12/4/2012 | 9:25:17 PM
re: MPLS Vendors Demo Fast Reroute 1) IP is not profitable
------------------------
Profitable or not, its here to stay. There is no technology to replace the services it provides to customers..

--------------------------------------

IP is a religion. It is not a business. Whether it make money or not does not matter.
==================================

No. I'm saying that the is a customer base out
there for services that can only be delivered by
IP. The internet and the consumers of internet
services over IP are not going to disappear.

Nearly all the airlines are unprofitable at this
point in time as well. But to conclude from that
that air travel is a "religion" and not a
"business"....or that people will stop flying is
absurd. The same analogy holds for IP.
----------
What is a IP service??

Please tell me..

I have no idea..
-----------------
Lets say your dream comes true and every IP router
is decomissioned. No more IP anything.

The internet ceases to exist. I guess we go
back to dialup and private line for all networking
between computers. Every connection to a server
anywhere is a call or a circuit setup. Every
connection involves charges and possibly charges
at both ends of the connection.

Unlike you, most people see a value to the
internet as it is. And don't see retreating back
into the bad old days as anything positive.








jamesbond 12/4/2012 | 9:25:17 PM
re: MPLS Vendors Demo Fast Reroute Thank you everybody for the pointers and comments.

Currently in the industry, how often is formalism used and how much do u value a person who has this skill.
----------------------------------

Most of the protocol related work being done
in the industry involves integration/porting etc.
Hardly anybody develops standardised protocols
from scratch, except ofcourse protocol vendors.
There are a hand few of these guys. Then again
what new heavyweight protocols are being
developed that need formal methods? None.

fiber_r_us 12/4/2012 | 9:25:17 PM
re: MPLS Vendors Demo Fast Reroute I am really tired of the "IP is not profitable" mantra! If IP services are unprofitable, why do the major service providers continue to sell the services and upgrade thier networks?? AT&T, Sprint, UUNET, Verizon, Level3, etc. all continue to offer and deploy new IP (or MPLS) based services. While companies sometimes do dumb things, I don't think all of these companies are continuing to offer a fundamentally non-profitable service in these tough economic times.

While many service providers are in trouble, this is mostly due to thier extensive build-out of thier physical and transport infrastructures during the bubble and the subsequent debt load that now hampers thier profitability. This has little to do with how many IP routers or MPLS switches they purchased. Indeed, almost all service providers have stated that they plan a migration to an MPLS infrastructure so that they may reduce thier overhead and try to become profitable once again.

Those that wish us to believe that the IP/MPLS services are unprofitable are undoubtedly trying to protect thier own vested interests in thier legacy technologies that are being threatened by the more cost effective IP/MPLS technologies.
techstud 12/4/2012 | 9:25:18 PM
re: MPLS Vendors Demo Fast Reroute C_junk - message 57:
<<
The new user-visible service is network-based L3 VPN. This
is extremely popular. It is definitely beginning to cut
into the market of enterprise networks based on FR/ATM PVCs.
That is because it offers certain value adds that the traditional
VC hub and spoke does not, e.g., any-to-any, plus since it is
IP the ability to use the TOS bits to provide differentiated
packet handling. >>

I am surprised how intelligent people like you on this list will come to wrong conclusions. So called, BGP/MPLS VPN model has very little to do with mpls, and more to with ugly hacking of BGP to carry vpn routes between PE routers in its multi-protocol attribute.

In the same model, you can send NULL label, and then use IP in IP tunnel or GRE tunnel or IPSEC tunnel between PE routers and also add a vpin-id equally effectively. I don't know why a user of VPN will care if the tunnel is MPLS or GRE or IP-in IP or whatever as long as it works.

A VPN will work fine with any tunneling protocol, presence of mpls in the network has no special benefits. Doesn't it add more compexity and operational cost?
optical_IP 12/4/2012 | 9:25:18 PM
re: MPLS Vendors Demo Fast Reroute hi

Thank you everybody for the pointers and comments.

Currently in the industry, how often is formalism used and how much do u value a person who has this skill.

optical_IP
jamesbond 12/4/2012 | 9:25:19 PM
re: MPLS Vendors Demo Fast Reroute There are/were several tools/languages designed
for this purpose -

LOTOS
ESTELLE
SDL

In fact I know of a LEX/YACC implementation of TCP. Do a search and you will find the links.
sgan201 12/4/2012 | 9:25:20 PM
re: MPLS Vendors Demo Fast Reroute Hi All,
http://www.nwfusion.com/news/2...

This for those interested in finding out how cheap FR service is.. This is with Mesh..
IP VPN service is around $2K per location...
sgan201 12/4/2012 | 9:25:20 PM
re: MPLS Vendors Demo Fast Reroute Hi cc_junk,
The most popular network based L3 VPN service is IP enabled FR or IP over ATM with VR..
Look at Savvis implementation: IP over ATM with VR..
AT&T IP enabled FR service is using FR switch to provide concentration and maybe RFC 2547 VPN to do the VPN..
Sprint PCS and WCOM is launching IP enabled FR service also..
Please noted that in IP enabled FR services MPLS RFC 2548 VPN is optional and in fact it is not as scalable as if you would by using a Cosine box...
Either, this service is a joke anyhow..
Total service revenue from IP VPN is less than one biilion dollars a year as compare to 12 billions dollar in FR..
Even if IP VPN double every year, it will take 3 to 4 years for it to surpass FR..

Have you check IP VPN pricing in real world??
It is around $2K per month..
With WCOM selling FR service at $236 per month and provide full mesh and better QOS and SLA guarantee, IP VPN does not stand a chance in USA..

cc_junk 12/4/2012 | 9:25:21 PM
re: MPLS Vendors Demo Fast Reroute dreamer101 said:
"MPLS is very useful of keeping your competitor investing
on useless stuff that no one will buy. Then, you can focused
on creating something marketable with little or no competition ..."

If MPLS was only about the traffic engineering part (including TE)
then I don't think it would survive. TE and FRR are not end
user-visible services. So they could easily be replaced by
alternative solutions.

The new user-visible service is network-based L3 VPN. This
is extremely popular. It is definitely beginning to cut
into the market of enterprise networks based on FR/ATM PVCs.
That is because it offers certain value adds that the traditional
VC hub and spoke does not, e.g., any-to-any, plus since it is
IP the ability to use the TOS bits to provide differentiated
packet handling.

L3 VPN's are being implemented using MPLS, although it need
not be. Just as FR is implemented using ATM. Nevertheless,
it is the implementation that has gained mind/market share,
so any alternative implementation won't get off the ground
because it will offer no sufficient benefit to switch midstream
this late in the game. Even though the L3 VPN user does not
see MPLS directly, the users and market view this as the MPLS
service/network.

Because of this new service, MPLS will not be going away.

Although L2 VPN is not a new service (FR, ATM VC services and
TLS have been around for years and mostly implemented over ATM),
the MPLS implementation allows those carriers/providers who
never offered those services to enter the market by converting
their IP network rather than build an ATM network (witness
Level 3). Unfortunately, it puts more competition and supply
into what has been a profitable market for carriers. But the
desire of those newer providers to get into a profitable new
market will also drive the deployment of MPLS switches.

Then there are the efforts of the incumbents who are studying
whether they can converge their data newtorks. They have all
attemped this to various degress using ATM as the multi-service
core. But for many the IP requirements and platforms (in terms
of switching capacity) have outstripped the ATM space [ATM was
meant to be the high speed data platform. Who would have thought
we could select from a large supply of OC48 and OC192 routing
platforms before the ATM space had a comparable offering. Just
shows you can solve any problem given enough investment of money
and people.] Some of these incumbents (see previous article on
Lightreading) apparently think they might be able to converge
networks finally using MPLS. If they justify it (or just do it
without justification) then it will provide additional MPLS switch
demand (the largest if they really follow through and migrate).

Because MPLS will be deployed to create both the new and old
services, then it is more likely that the MPLS TE and FRR will
also be widely used.
sgan201 12/4/2012 | 9:25:22 PM
re: MPLS Vendors Demo Fast Reroute Hi Skeptic,
Thanks for clarifying the truth which I have been trying to get the IP folks to say out loud.

1) IP is not profitable
------------------------
Profitable or not, its here to stay. There is no technology to replace the services it provides to customers..

--------------------------------------

IP is a religion. It is not a business. Whether it make money or not does not matter.

So, here it goes..
All SP that do not wish to make money, they can go ahead and do IP and bankrupt. And, they did..


What is a IP service??

Please tell me..

I have no idea..

What is the irreplaceble IP service that SP provide to customer??

Does it exists??

What is the revenue figure??

By the way, the ATM switch revenue figure at Lucent, MArconi, Alcatel, and NT is still many times higher than Juniper core routers..
h 12/4/2012 | 9:25:23 PM
re: MPLS Vendors Demo Fast Reroute One of the best known works in the academic environment for protocol verification is a tool called SPIN by bell-labs. Its been around for 20 years with continuous evolution. It does lot of checking based on formal methods and I am not knowledgable to say exactly what they are - at a high level they do include tests for deadlock, livelock, etc of protocols and they are provable guarantees for your protocol as defined by you.

It is not however there to help u architect ur protocol nor has anything to do with software design or software testing. I do recall instances of SPIn being used to test some real code and it is pretty impressive. check it out
http://spinroot.com/spin/whati...
Also check the associated book with spin titled
Design and Validation of Computer Protocols by Gerard J. Holzmann
an online version of the book is also available at
http://spinroot.com/spin/Doc/p...

And there are SPIN workshops held yearly and u can check its proceedings to see if there already work related to the context in which u are looking.

Hope these help
dljvjbsl 12/4/2012 | 9:25:24 PM
re: MPLS Vendors Demo Fast Reroute TTCN is used for functional test. It is not used for the formal logical verification of a design but as a means of doing a state space fucntional test. In a realisitc protocol, the number of states explodes astronomically and full state space tests are not very useful.
dljvjbsl 12/4/2012 | 9:25:25 PM
re: MPLS Vendors Demo Fast Reroute Something as formal as telephone call control systems have been formalized. There are two maindifficulities - one is that formalisms are brittle and much work has to be done even for the simplest change in the system and secondly that it is difficult to get anything more than trivial results out of the test. it is difficult to know what to verify mathematically and so bugs can reamin in the system becasue no one knew to look for them.

Pure formal methods have not lived up to the hopes that people placed in them.

As a result of this, pure formal techniques are being eschewed in favor of what has been called appropriate formality. The system is modelled and areas of interest are identified and given the full logical treatment. In other areas human level techniques of software engineering are uswed to create good dsigns. Design patterns, modular design, developed test and communcation, etc are all used.

Teh msot important thing that has been realized in this is that 'correctness' is only part of a good desing. Evolvlability and stability in the face of local and distant errors are just as important and cannot be proved mathematically.
BGP_OSPF_ISIS_RIP 12/4/2012 | 9:25:25 PM
re: MPLS Vendors Demo Fast Reroute I've heard about this tool but haven't used it myself. It's called TTCN (Tree and Tubular Combined Notation) and was developed as part of a thesis at University of Ottawa. http://www.itu.org has more info.

Good luck.
skeptic 12/4/2012 | 9:25:26 PM
re: MPLS Vendors Demo Fast Reroute what are the formal methods used while designing new protocols / specs. Like VLSI CAD tools, are there any tools to design and verify protocols.
-------------------------

In the industry, there are few formal methods
used. In the research/academic world, they tend
to use simulation packages (mathmatical) to
demonstrate the stability of traffic under
a particular model.

But the protocols themselves tend to be too open
ended for easy simulation....or at the other
extreme are too trivial for formal verification.

The other problem is that most of the stability
problems associated with routing protocols are
issues with how they were implemented rather than
isuses with the stability of the standard. The
IETF standards are vague and in OSPF/ISIS/BGP/
RSVP/etc, the difficult thing is designing the
software to meet the standard and deliver
reasonable performance.

Most of the best work (for example in BGP) was
done by observations/measurements in live
networks. Problems are identified and then
fixed.

skeptic 12/4/2012 | 9:25:26 PM
re: MPLS Vendors Demo Fast Reroute "The truth will set you free"

1) IP is not profitble
------------------------
Profitable or not, its here to stay. There
is no technology to replace the services it
provides to customers.

------------------------
2) How many of those boxes actually has MPLS turned on??
-------------------------
Quite a few actually.

==========================
3) How many of those boxes actually has RSVP+TE turned on??
--------------------------
Nearly all. Because I dont know of anyone who
does static MPLS labels by choice.


4) Have you check the latest revenue figure of Juniper??
----------------
Have you checked the situation at lucent, nortel,
alcatel and Marconi?

----------------
optical_IP 12/4/2012 | 9:25:26 PM
re: MPLS Vendors Demo Fast Reroute hi,

this may be out of context of this thread, but anyway ......


what are the formal methods used while designing new protocols / specs. Like VLSI CAD tools, are there any tools to design and verify protocols.

how do u ensure that the new protocol works correctly under varied conditions ? (verification)
(BGP route flapping .... error in initial spec ? )

any insights would help.

optical_IP
sgan201 12/4/2012 | 9:25:31 PM
re: MPLS Vendors Demo Fast Reroute Hi lob,
After all said and done, if you want to sell something now. Do APS on the MPLS switch and you can sell it now.

In networking, we have legacy equipments.
Class 5 switches that are installed out there will drop your voice calls periodif it is more than a few seconds interruption.

Are you going to tell your customer that you cannot use my equipment unless you swapped out all your class 5 switches??

Even voice is not a problem. You still have SNA to deal with. SNA has a 2 seconds limit. You do want to withdraw your money from the ATM machine??
SNA is the most important/mission critical data comm protocol in USA.

Don't be religious. Solve the customer's networking problem in the most effcient and cost effective way.. Give them the choices?? Proven, simplicity, and cheap is always the better way. KISS..
sgan201 12/4/2012 | 9:25:31 PM
re: MPLS Vendors Demo Fast Reroute Hi Tony,
If I would to read through you whole e-mail and filterred out all the noise. The summary is we are doing MPLS FRR because some customer want that.
However, from out discussion, it seem like those customer are not very good network architect/engineer.

sgan201 12/4/2012 | 9:25:31 PM
re: MPLS Vendors Demo Fast Reroute Hi,
At which planet, MPLS based service is easier and has good IP QOS?? It is sure not on this planet Earth.

The last time I look, the only service that fit your description is either FR or IP over FR..
WCOM FR service is $236 per month..
Which/whose MPLS service is cheaper than that and has QOS/Jitter guarantee..

Quote me a real service and not a science project...
sgan201 12/4/2012 | 9:25:31 PM
re: MPLS Vendors Demo Fast Reroute Hi imref,
"The truth will set you free"

1) IP is not profitble

2) How many of those boxes actually has MPLS turned on??

3) How many of those boxes actually has RSVP+TE turned on??

4) Have you check the latest revenue figure of Juniper??

5) Check the revenue figure of Level 3??

6) Without FR/ATM, MPLS is toasted
Tech_Worker 12/4/2012 | 9:25:32 PM
re: MPLS Vendors Demo Fast Reroute A person claiming to be Tony Li said:

TL: <<yes, a="" able="" and="" as="" be="" break="" call="" calls="" customer="" engineered="" failure.<br="" i="" link="" lsps="" mentioned,="" not="" or="" over="" put="" requirement="" router="" the="" there="" to="" traffic="" voice="" was="" when="">
Of course, my requirement is actually simpler: how do I sell routers to people who want circuit switching?>>

TW: [With the addition of mechanisms like FRR (that require more computing cycles & state information in the network), aren't we getting closer to circuit switching?

Aren't we saying: "The way forward for QOS is to introduce more cycles & states in the network?"

How about saying we are trying to retrofit the current IP based packet switching infrastructure to try and address the constraints of communicating bits in the space-time domain like circuit switching is good at?
]
</yes,>
lob 12/4/2012 | 9:25:32 PM
re: MPLS Vendors Demo Fast Reroute > Of course, my requirement is actually simpler:
>how do I sell routers to people who want circuit
>switching?

Tough. This is what this FRR thingie really is about, if we drop pretenses :)

I think sub-50 ms service restoration is quite doable (using few cheap tricks I outlined) even with plain SPF, in any realistically-sized network. With modern CPUs you can do an awful lot in 5 ms, particularly if all preparations are done in advance. (On a back of envelope... 5ms = 2e2 Hz, P-4 clock is 2e9 Hz, so you have 10 million instruction cycles to process link-state update - this does not strike me as a tight budget :)

Actually, does anyone really care about 50ms restoration? There's no reason for voice calls to be dropped if there's less than few seconds interruption - paying public tolerates that just fine (as cell companies taught everyone).

My guess is that voice people simply want to kill the entire integration thingie by injection of nonsensical requirements - so the data folks will leave them alone so they can continue worshipping their beloved mammoth switches. I'm inclined to think that they are right and that it does not make sense to replace SS-7 infrastructure which is already sunk capital.

The grass roots VoIP will kill them eventually, so nobody would have to worry about integration and compatibility with Bellcore-themed acid trip notes.
lob 12/4/2012 | 9:25:33 PM
re: MPLS Vendors Demo Fast Reroute Tony --

> I agree that debouncing the circuit is a good
> idea, but again, that's part of the detection
> time.

If detection time is order of magnitude higher than the reaction time what use is to spend inordinate efforts on improving reaction? Baaaad engineering :)

"Using MPLS sensibly" - ghmm. People are hell-bent on using it insensibly. Like using MPLS to signal optical layer to switch things around :)

[FRR doing non-optimal routes] --

>Absolutely. Again, the point of FRR isn't to be >optimal. It's to recover. You want optimal, you >then compute a backup path and do the >make-before-break dance.

What I was saying is that it is possible to do path-cost optimal routing in, essentially, the same amount of time (as seen by end-user) it takes to do FRR. So, what exactly is the gain in doing separate recovery?

Note that I used pure SPF specifically to simplify argument. The same argument applies to SPF augmented with TE features as well.

> Sorry, but you are trying to solve a
> completely different problem.

No, I'm asking if there's a problem which needs the solution called FRR. Sorry for being critical a-posteriori, there's so much noise that I didn't pay this one any attention until I saw your name. Anyway, knowing which windmills to attack is sometimes useful :)

And no, you didn't guess right: I couldn't make my hair in mohawk, it is just too long for that :)
Tony Li 12/4/2012 | 9:25:33 PM
re: MPLS Vendors Demo Fast Reroute Dear Long Hair,

----------------
If detection time is order of magnitude higher than the reaction time what use is to spend inordinate efforts on improving reaction? Baaaad engineering :)
----------------

Well, you should understand that the requirement from the customers at the time was to provide a way to restore service in a mesh network that was competitive with SONET. And SONET, of course, has the exact same issues with the debouncing of the circuit. And given that the customer wanted to use LSPs already, adding FRR was hardly inordinate effort.
---------------

"Using MPLS sensibly" - ghmm. People are hell-bent on using it insensibly. Like using MPLS to signal optical layer to switch things around :)

---------------
Well, I have to agree, the hype and the misinformation is rampant and cannot be replaced by sane engineering. So as always, keep your hype filter set on "high". BTW, people are NOT talking about using MPLS to signal optical changes. Instead, they are talking about extending the routing and signalling from the TE stuff into the optical domain. That's bringing over the routing and signalling work associated with MPLS, but MPLS itself would NOT be involved, in the sense that there wouldn't need to be labels on the lambda.
----------------

>Absolutely. Again, the point of FRR isn't to be >optimal. It's to recover. You want optimal, you >then compute a backup path and do the >make-before-break dance.

What I was saying is that it is possible to do path-cost optimal routing in, essentially, the same amount of time (as seen by end-user) it takes to do FRR. So, what exactly is the gain in doing separate recovery?
----------------

Ok, I'll try again. First, the whole point of doing this is NOT to do something that is path-cost optimal. The point was to take traffic that already needed to be traffic engineered and then to restore that service. Traffic engineering is, by definition, more general than path-cost optimal and frequently trades off optimality on one path to achieve overall optimality for the network.

Second, to restore service on an LSP in less than 50ms is non-trivial without FRR. Look at more of the details of what would happen with a link state IGP. From the point of failure, updated link state advertisements have to flow back towards the source up to the point where there is a branch point onto the optimal path. And these packets have to travel in a hop by hop fashion, touching each processor along the way. And then SPF has to happen and routes have to get installed. Only then does traffic continue to flow. How do you do this in the given time budget when the propagation time of the link state message could alone be as much as your 50ms?

With FRR, there is one change per LSP that needs to be made, just upstream of the break. And then service resumes, albeit along a sub-optimal path.
----------------

> Sorry, but you are trying to solve a
> completely different problem.

No, I'm asking if there's a problem which needs the solution called FRR. Sorry for being critical a-posteriori, there's so much noise that I didn't pay this one any attention until I saw your name. Anyway, knowing which windmills to attack is sometimes useful :)
---------------

Yes, as I mentioned, the customer requirement was to be able to put voice calls over traffic engineered LSPs and not break the call when there was a link or router failure.

Of course, my requirement is actually simpler: how do I sell routers to people who want circuit switching?
----------------

And no, you didn't guess right: I couldn't make my hair in mohawk, it is just too long for that :)

----------------

Well, then you must be Mr. Prius.

Tony
imref 12/4/2012 | 9:25:34 PM
re: MPLS Vendors Demo Fast Reroute Show me some real significant MPLS switch purchase by any SP..

---
i think there are a few carriers out there running Cisco 12000 and Juniper M160s.
teng100 12/4/2012 | 9:25:34 PM
re: MPLS Vendors Demo Fast Reroute ""
Enterprises that I speak with are looking at MPLS-based service for the following reasons:
- easier & cheaper any-to-any connectivity
- IP QoS support from the SP
""
You can also talk to the SP who provide this
kind of service to see when they can make the ROI?
based on this model. Following what customer wants
may not always make a business case.

imref 12/4/2012 | 9:25:35 PM
re: MPLS Vendors Demo Fast Reroute Why should a SP work so hard to earn less money??
Please noted the number that you quote "thousand of customers".. FR has millions of customers and billions of revenues..
------

You could have made the same arguments to justify retaining X.25 services instead of spending money to build out FR/ATM networks.

Enterprises that I speak with are looking at MPLS-based service for the following reasons:
- easier & cheaper any-to-any connectivity
- IP QoS support from the SP

If FR/ATM SVCs had taken off, perhaps MPLS would be irrelevant.
Tony Li 12/4/2012 | 9:25:35 PM
re: MPLS Vendors Demo Fast Reroute >In case you missed the point, the speed of light
>propagation delays are completely irrelevant in
>fast reroute because there is NO signalling from
>the hop before the failure back to the head end
>of the LSP.

I'm afraid you are missing the point this time, Tony. I'll try to explain why.

First of all, the LOS is not generated out of thin air. To reroute LSPs you need indication that your _transmitting_ line is broken. This means that in the worst case you get it after RTT of the line plus time between remote fault signal bits.
------------

That's correct, but that part of the time is considered part of failure detection. That time is not included in SONET's 50ms, so we have a difference in times that aren't part of the budget anyhow.

-----------
[ This is my initial objection to sloppy reporting misleadingly comparing 5-10ms with SONET's 50ms - FRR is NOT in any way faster than APS ].
-----------

I disagree AND don't care. The budget is 50ms, faster is basically irrelevant.

------------
Then, as any network operator worth its salt knows, the one-off outright faults are not a problem - those happen seldom, and do not reduce network availability much. The real problem is glitches and bouncing circuits. That is why most routers installed out there are configured to _ignore_ LOS. In the real life it is quite preferrable to ignore short glitches and simply drop packets on the floor rather than change direction of packet flows. That makes "fast" rerouting pretty much irrelevant - a typical timer for detection of link failure is 30 sec. Admittedly, it is because of high cost of rerouting and could be reduced to 300-500 msec, but that anyway swamps the difference between "fast" rerouting and plain old SPF.
----------------


I agree that debouncing the circuit is a good idea, but again, that's part of the detection time.


-----------------
My final (and most serious) objection to FRR is that it is built on a fundamentally flawed premise that total convergence time matters.

In the reality routing updates fly over wires at the exactly same speed as the data packets. So, lets assume (for simplicity) that routing algorithm computes instaneously. In this case the routing updates go in a wave _ahead_ of any actual rerouted data - so you end up with optimally routed paths just in time to receive data flows.
-------------

Yes, but the goal wasn't to have optimally routed paths. It was to have traffic engineered paths that rerouted in time frames that didn't break phone calls. Just running SPF faster wouldn't help. And plus, you're ignoring the time that it takes for the failure information to flow from the
failure back to the point where the rerouting needs to happen.

------------
Let's compare that with FRR. You have to precompute different scenarios. You need to update LSP tables to line cards (and those LSP tables are O(n^2), not linear to the network size!
-------------


Only if you're abusing MPLS. If you're living a bit more sensibly, you're using something that's
linear or close to it.

Yes, I know that there are some people out there who think that MPLS was supposed to replace full mesh ATM. I hope that those folks are in the minority, but I also know that they've made up their minds and won't understand the scalability issues.
--------------

In the end you end up with manifestly non-optimal paths (it is very easy to draw examples of realistic topologies where FRR would create backup LSPs dragging packets back - creating what is used to be called persistent routing loops). Note that routing loops in SPF are transient.

----------------

Absolutely. Again, the point of FRR isn't to be optimal. It's to recover. You want optimal, you then compute a backup path and do the make-before-break dance.

----------------

Sorry, Tony, the problem has a much simpler solution, not requiring any new-fangled and certain to be buggy technology.

----------------

Sorry, but you are trying to solve a completely different problem.

-----------------

By now, I guess, you can guess who I am :)

------------------

You be Mohawk-Boy. ;-)

Tony
sgan201 12/4/2012 | 9:25:35 PM
re: MPLS Vendors Demo Fast Reroute Hi,
Which year the SP will buy enough MPLS equipment to keep any MPLS switch start up alive??
Care to forecast/predict....
If it is 2005 or 2006, then, it is too late...

Talk is cheap..

Show me some real significant MPLS switch purchase by any SP..
lob 12/4/2012 | 9:25:37 PM
re: MPLS Vendors Demo Fast Reroute Tony --

>In case you missed the point, the speed of light
>propagation delays are completely irrelevant in
>fast reroute because there is NO signalling from
>the hop before the failure back to the head end
>of the LSP.

I'm afraid you are missing the point this time, Tony. I'll try to explain why.

First of all, the LOS is not generated out of thin air. To reroute LSPs you need indication that your _transmitting_ line is broken. This means that in the worst case you get it after RTT of the line plus time between remote fault signal bits.

[ This is my initial objection to sloppy reporting misleadingly comparing 5-10ms with SONET's 50ms - FRR is NOT in any way faster than APS ].

Then, as any network operator worth its salt knows, the one-off outright faults are not a problem - those happen seldom, and do not reduce network availability much. The real problem is glitches and bouncing circuits. That is why most routers installed out there are configured to _ignore_ LOS. In the real life it is quite preferrable to ignore short glitches and simply drop packets on the floor rather than change direction of packet flows. That makes "fast" rerouting pretty much irrelevant - a typical timer for detection of link failure is 30 sec. Admittedly, it is because of high cost of rerouting and could be reduced to 300-500 msec, but that anyway swamps the difference between "fast" rerouting and plain old SPF.

My final (and most serious) objection to FRR is that it is built on a fundamentally flawed premise that total convergence time matters.

In the reality routing updates fly over wires at the exactly same speed as the data packets. So, lets assume (for simplicity) that routing algorithm computes instaneously. In this case the routing updates go in a wave _ahead_ of any actual rerouted data - so you end up with optimally routed paths just in time to receive data flows.

Therefore, non-locality of SPF does not matter,
and locality is NOT advantageous.

Now, the objection is, of course, that it is hard to do routing computations fast. Actually, running Dijkstra after receiving link-state update is not necessary - you can always keep precomputed paths for all possible single-link faults in memory. The really expensive part is convolving exterior and new interior routing info together and loading it into routing engines.
This step is only needed because hardware is designed by people who never heard about iBGP hack - otherwise they'd support two-step IP address resolution in hardware - a complete no-brainer.

In that case, the only action needed to effect SPF route change on receipt of link-status notification is to load small {O(n) from network size} interior routing table into line cards. In fact, those tables can be pre-loaded. With a little more effort, you can prevent loss of packets after discovery of link failure altogether by saving them in buffers (which are more than large enough to accomodate inflow for the time needed for a routing algorithm to work).

Let's compare that with FRR. You have to precompute different scenarios. You need to update LSP tables to line cards (and those LSP tables are O(n^2), not linear to the network size! Ironically, by its very nature FRR makes those tables even bigger - making pre-loading them to line cards even less practical).

In the end you end up with manifestly non-optimal paths (it is very easy to draw examples of realistic topologies where FRR would create backup LSPs dragging packets back - creating what is used to be called persistent routing loops). Note that routing loops in SPF are transient.

So... we have a technique which is NOT faster on hardware of similar coplexity/cost, prone to creation of persistent routing loops, having O(n^2) (instead of O(n)) scaling properties, and requiring an additional layer of routing software (aka MPLS). All of that to compensate for deficient design of hardware and routing software.

Smells like tag switching all over again. Sorry, Tony, the problem has a much simpler solution, not requiring any new-fangled and certain to be buggy technology.

By now, I guess, you can guess who I am :)
supplyside 12/4/2012 | 9:25:38 PM
re: MPLS Vendors Demo Fast Reroute What is this? So much fuss over such a simple 'technology'. Come on folks, this is not rocket science. All you have to do when a link fails is change the outgoing interface for the appropriate incoming labels, and also add another label to pushed onto the label stack as per the pre configured reroute LSP. The main issue here is a configuration issue, that the network operator configure the reroute LSPs properly so as not to make scrambled eggs of the network.

All big fuss about nothing.

-supplyside
AAL5 12/4/2012 | 9:25:39 PM
re: MPLS Vendors Demo Fast Reroute If some people think service providers are not asking for MPLS enabled switches/routers they are living in a delusional world.

The fact is that major SPs have said that they see their ATM/FR networks transitioning into using an MPLS infrastructure for L2 and L3 VPNs.

Using MPLS gives them the following benefits:

-Takes BGP out of the core, reducing routing convergence/routing flaps (i.e. enhances routing stability for core data network)

-Allows SPs to provide L3VPN services (revenue generating)

-Allows SPs to provide L2VPN services to customers based on their existing IP network (revenue generating and reduces capex)

-Allows Opex for FR/ATM/IP networks to be amalgamated into one network that provide regular IP L2VN and L3VPN services.


All of you that do not want to chase after this business great, it means there is less competition for those of us that do wish to provide equipment to meet the requirements of customers.

AAL5
skeptic 12/4/2012 | 9:25:41 PM
re: MPLS Vendors Demo Fast Reroute Do you have any study that gives me ROI for MPLS?
What are the "benefits" nobody tells me except presenting marketing cliches?
----------------------

Sure.


1) If you have an IP backbone already, the
incremental capital cost in equipment is near
zero. There are operational expenses, but they
depend on what you do with MPLS. At a minimum,
it gives you a large toolbox of features for
operationally improving IGP routing and for
opaque tunneling at far lower OPEX cost than
the technical alternatives.


2) Calculate how much capex and opex are
associated with running IP/ATM/FR as three
seperate services with dedicated equipment.
Then start calculating the savings over the
long term associated with the elimination of
ATM/FR dedicated equipment from your network.

3) Ask yourself if the margins in private-line
business which subsidize ATM equipment are
sustainable over the long term. Start looking
at your customers and you will find out that most
of them are not Real-time fine-grained QOS ATM
customers. They just want private line service.


----------------------
Please provide, if you can, any economic justification for mpls how it HELPS carriers make more money by offering any NEW revenue services which they couldn't offer without the use of existing or alternate technologies?
----------------------------
"services" was never the motivation for MPLS.
Nor was trying to give service providers economic
models for their business. Vendors sell equipment
based on what customers ask for. You can't
run to the vendors and demand that they tell you
how to run your business and make money.
------------------------
As a tunneling protocol, MPLS is useless it neither has in-built security or authentication like IPSEC, nor control channels maintenance, and other tunnel management issues like L2TP as part of its definition.
------------------------

People can't have it both ways. On one side,
everyone complains that MPLS has too much it
or is too complex.....on the other they want
it to duplicate the functionality of IPSEC
and L2TP!

The value of MPLS tunneling is that its very
generic and very simple. If you want security,
encrypt the traffic you are tunneling. MPLS
provides a generic tunneling scheme within the
provider that many people find very useful.
-------------------
I don't any other use of MPLS except as a vpn tunneling protocol (remember, mpls was designed as a packet forwarding mecahnism not as a tunneling protocol). so to transform a packet forwarding mecahnism at a node into a wide area tunneling protocol, you add mindless layers after layers of complex tunnel management mechanisms, crankback schemes and fault recovery of type this and type that (most of them straight lift from circuit switch/atm literature).
-------------------------------

Honestly, have you ever used MPLS? And by
MPLS, I mean on a router with RSVP-TE signaling
and TE extensions. Tell me what these
"mindless layers" are, because I'm not sure
what your talking about.

-------------------------------
MPLS requires IP, and on IP network a simple tunneling protocol like L2TP can transport any data whatever mpls can. It is just that mpls has been marketed better. There is no money in simple solutions for Cisco or Juniper.
-------------------------------

Why don't you go talk to the people at the
major national service providers who demanded
MPLS from the vendors. They wanted it and they
are using it now. Obviously, they had problems
that needed a better solution than L2TP.











beowulf888 12/4/2012 | 9:25:41 PM
re: MPLS Vendors Demo Fast Reroute Many thanks to dreamer101, rationalMind and planoboy for their detailed analysis of FRR and APS. Dreamer: I admit I wasn't considering BLSR configurations -- my experience with SONET is largely limited to UPSR installations on big corporate networks (back when ATM was the protocol that was going to kick TCP/IP's butt).

Frankly, I'm still digesting this information (and next I'll be cracking open my dusty SONET text book). I have to say that the discussion on the Lightreading lists is excellent when we stay away from religious wars!

cheers to all...
--Beo
sgan201 12/4/2012 | 9:25:42 PM
re: MPLS Vendors Demo Fast Reroute Hi,
Talk is cheap. Real equipment order and real service revenue is the key.. How much MPLS service revenue do those SP get??
I will bet it is far less than one billion dollars for all service providers.
By the way, the total service revenue for IP VPN is less than one billion dollar (MPLS + Network based + CPE bsaed)..
And, the margin on those IP VPN service is much lower than FR/ATM. I read somewhere it is 30-45% for FR versus 10-15% for IP VPN..
Why should a SP work so hard to earn less money??
Please noted the number that you quote "thousand of customers".. FR has millions of customers and billions of revenues..
ImAClone 12/4/2012 | 9:25:45 PM
re: MPLS Vendors Demo Fast Reroute The good thing about the test is that there were 3 vendors who interoperated, given the history of the FRR draft(s).

The not so big deal is that failover should scale locally, and the test didn't show that the boxes could do so. Most of these boxes support 16K or more LSPs. Can as many LSPs fail over, all of them in generally sub-50 msec times? That would be more interesting, albeit implementation-related information.

-clone
imref 12/4/2012 | 9:25:45 PM
re: MPLS Vendors Demo Fast Reroute I'd recommend you talk to folks such as BT, AT&T, Infonet, Equant and others who have deployed, and are actively selling MPLS-VPN services to thousands of customers.

As far as new services - check out recent announcements from Cisco, Redback and others on new service opportunities enabled by MPLS (e.g. address pooling, managed Internet gateways, multicast, etc.)

L2TP is not well suited for large, meshy type deployments and has serious scalability issues in such applications since there is no way to combine multiple tunnels using something like MPLS label stacking. It also doesn't inherently have the ability to perform traffic classification based on IP header information.

Plenty of people agree with your philosophy, but every major carrier in the U.S. - except for Sprint and Savvis - are building MPLS-based services. Perhaps everyone in those companies is wrong and you are right. We'll see in due time. If I were you, I'd spend a bit of time speaking with the senior folks in those organizations and trying to gain an understanding into what's driving their MPLS deployments.
imref 12/4/2012 | 9:25:46 PM
re: MPLS Vendors Demo Fast Reroute On the contrary, MPLS got complicated in the past few years. MPLS initially was meant to ease and simplify routing but kept extending its architecture into a layer of complications. It should be standardized and split into implementations so customers can tailor it according to their needs, such as, vanilla MPLS, TE E-LSP MPLS vs TE L-LSP MPLS, BGP/VPN MPLS, etc..

-------
that's exactly what the IETF sub-IP working groups are doing.
techstud 12/4/2012 | 9:25:47 PM
re: MPLS Vendors Demo Fast Reroute imref:

<<mpls added="" additional="" and="" benefit="" complexity,="" continues="" cost="" deployed="" enabling="" have="" in="" ip-based="" is="" it.="" majority="" network="" of="" offer="" providers="" real="" reducing="" service="" services="" services,="" the="" to="" transport="" value="" vast="" wan="" which="" why="">>

Do you have any study that gives me ROI for MPLS?
What are the "benefits" nobody tells me except presenting marketing cliches?

Please provide, if you can, any economic justification for mpls how it HELPS carriers make more money by offering any NEW revenue services which they couldn't offer without the use of existing or alternate technologies? Which US carriers have ALREADY DEPLOYED it over wide area? Deployed in Finland who cares? All they need is 2 GSR 12000 in Helsinki to call it a nationwide network :-). Doesn't it tell you something about that US market is very different than Europe and Asia?

As a tunneling protocol, MPLS is useless it neither has in-built security or authentication like IPSEC, nor control channels maintenance, and other tunnel management issues like L2TP as part of its definition. I don't any other use of MPLS except as a vpn tunneling protocol (remember, mpls was designed as a packet forwarding mecahnism not as a tunneling protocol). so to transform a packet forwarding mecahnism at a node into a wide area tunneling protocol, you add mindless layers after layers of complex tunnel management mechanisms, crankback schemes and fault recovery of type this and type that (most of them straight lift from circuit switch/atm literature). So what is great about it? Doesn't MPLS look like ATM with a variable cell size?

MPLS requires IP, and on IP network a simple tunneling protocol like L2TP can transport any data whatever mpls can. It is just that mpls has been marketed better. There is no money in simple solutions for Cisco or Juniper.</mpls>
sgan201 12/4/2012 | 9:25:49 PM
re: MPLS Vendors Demo Fast Reroute Hi RationalMind,
I was saying that in the case MPLS FRR (not Sonet APS) and only the RX cable get cut, the RX side has to signal the other side (TX side) so that the MPLS switch know that the port has gone LOS..
If the MPLS switch do not know the port had gone LOS, it will never initiate FRR to begin with..
woodstock 12/4/2012 | 9:25:50 PM
re: MPLS Vendors Demo Fast Reroute <<
MPLS continues to offer real benefit in reducing network cost and complexity, and in enabling IP-based WAN transport services and additional value added services..
>>

On the contrary, MPLS got complicated in the past few years. MPLS initially was meant to ease and simplify routing but kept extending its architecture into a layer of complications. It should be standardized and split into implementations so customers can tailor it according to their needs, such as, vanilla MPLS, TE E-LSP MPLS vs TE L-LSP MPLS, BGP/VPN MPLS, etc..
planoboy 12/4/2012 | 9:25:51 PM
re: MPLS Vendors Demo Fast Reroute Since I have done a lot of protection switching work here goes my BS :-) -
The minor issue of AIS - SONET AIS (path and line) is sent downstream and RDI is sent back to the Far end. However as someone mentioned we use SONET K1/K2 bytes to effect protection switching in Linear(1:1)/BLSR architectures.

Unidirectional versus Bidirectional - UPSR (Unidirectional Path Switched Rings) and Linear (1+1 uni) architectures switch only at the near end (end detecting the failure - LOS/LOF/AIS) and do not require signalling to the far end.
From my experience, some customers did not like this for the following reason - If the near end was selecting protect traffic and the far end was selecting working traffic, the craftperson does not know which card to pull out and claim it is bad !. Of course, he would later write up a problem report and my manager would be screaming at me. On the other hand if it was bidirectional, the far end and the near end would both be selecting working or protect. Those damn LEDs would show that !.
Bidirectional - is obviously harder to implement - requires processing of K bytes. The benefit is in the extra traffic or Protect Channel Access(as one prominent vendor would call it :-) since the head end is not bridging the working and protect channels.

IS the FRR good ? Sure, if unidirectional is all they require and they can live with traffic being bridged.
rationalMind 12/4/2012 | 9:25:51 PM
re: MPLS Vendors Demo Fast Reroute No, I am not wrong!!! You just dont understand that SONET APS works in 2 mode - uni-directional and bi-directional. Read post 22 for detailed explaination.
imref 12/4/2012 | 9:25:52 PM
re: MPLS Vendors Demo Fast Reroute >RBOC want to buy ATM switch.
>There are only 2 next gen ATM switch start up: >Wavesmith and Equipe..
>Meanwhile everyone waste their money and effort >on MPLS...
--------------

In reality, every RBOC has an RFP on the street for MPLS, they are all considering MPLS for WAN services if they get approval to go beyond their home areas. This is why every vendor under the sun is pushing their MPLS solutions.

In reality, just about every new network being built in places such as Asia and Europe is based on MPLS.

In reality, we're long past the discussion of whether or not MPLS is a useful technology, the market has spoken.

The folks arguing that MPLS is bad because it violates the Internet architecture remind me a bit of those who argued that everything on the Internet should be free. They are following a principle that is no longer relevant, and will lead them to the unemployment line.

Service and technology solution selection should be driven by what will generate the most revenue, or minimize operational costs, period.

MPLS continues to offer real benefit in reducing network cost and complexity, and in enabling IP-based WAN transport services and additional value added services, which is why the vast majority of service providers have deployed it.

Now, can we finally move past the "MPLS is evil" discussion?
sgan201 12/4/2012 | 9:25:52 PM
re: MPLS Vendors Demo Fast Reroute Hi rationalMind,
You are wrong!!!
My statement has to do with how the optical port on both side of the cable detect that the port is down/LOS/bad.. It is a layer 1/2 issue fault detection issue. It has nothing to do with being multicast/simplex and so on..

Think of it this way..
In order for MPLS FRR to work, both boxes at the end of the cable has to detect the port is bad and switch the LSP to a different port. The only way for some port to detect an error is when it did not receive anything.. When RX cable get cut, box on the RX side need to send something to tell the other port on the TX side that the port is bad. And then, both box switch the LSP to the alternate route/port..

Imagine what happen if only one end of the cable/port switch and the other side did not switch??

MPLS LSP is simplex and this is even worst..
The sender has to switch the LSP but only the RX side can tell him whether the PORT/cable is bad..
Prizm 12/4/2012 | 9:25:54 PM
re: MPLS Vendors Demo Fast Reroute The proper test is to only cut the RX cable. The RX side will detect the failure. It will generate Sonet AIS to the TX side on the same port.

Bzzzz! Sorry, wrong answer.
AIS is only propagated in the same direction as the failure. This does not apply in this case since the router is Path Termination Equipment.
What the router should do is send RDI (remote defect indication) on its Tx port.
rationalMind 12/4/2012 | 9:25:54 PM
re: MPLS Vendors Demo Fast Reroute Typo Alert:

I wrote:
So, I think dreamer's criticism are not irrelavant

Should be:
So, I think dreamer's criticisms are irrelavant
spotlight 12/4/2012 | 9:25:54 PM
re: MPLS Vendors Demo Fast Reroute
AIS is only propagated in the same direction as the failure. This does not apply in this case since the router is Path Termination Equipment.
What the router should do is send RDI (remote defect indication) on its Tx port.

=====================================

If you send the AIS in the same direction it is Uni directional switching and if AIS is sent on the Tx port it is bidirectional switching

About Multicasting, How do you think Extra Traffic is provided ....
rationalMind 12/4/2012 | 9:25:55 PM
re: MPLS Vendors Demo Fast Reroute beowulf888 wrote:
dreamer:
In the scenario you presented, how does SONET ensure that the failover between ports on both switches happens at the same time? I guess I just don't see how there is any difference between MPLS FRR and SONET APS. Can anyone explain?

best regards,
--Beo

dreamer101 wrote:
"The proper test is to only cut the RX cable. The RX side will detect the failure. It will generate Sonet AIS to the TX side on the same port. Depending on the length of the TX cable, the other side may take a while to receive AIS. Obviously, if the TX cable is a few thousand miles long, it will take a while to switch over. How does MPLS FRR ensure that the port is switchover on both side at about the same time???"
---------------------------------------------------------

FRR and APS are very different in a number of ways. For one FRR doesn't involve multicasting traffic on both the working and protect lines while APS does. And this point is moot to dreamer's criticism. Also, SONET is bi-directional by nature while MPLS is not (I know you can signal bi-directional LSPs). So, I think dreamer's criticism are not irrelavant as you are not comparing apples to apples. Also, uni-directional SONET APS (I think more widely used than bi-directional APS) doesn't behave the way dreamer described.

sgan201 12/4/2012 | 9:25:56 PM
re: MPLS Vendors Demo Fast Reroute Hi,
Read GR-253 or any basic Sonet book. Sonet APS has K1 and K2 bytes that hand shake and synchronize this..
sgan201 12/4/2012 | 9:25:56 PM
re: MPLS Vendors Demo Fast Reroute Hi tera,
There is a big difference between asking for something and actually willing to buy it..
Who has actually pay big bucks for MPLS equipment??
Don't you know by now designing equipment by RFP is a sure way to failure??
beowulf888 12/4/2012 | 9:25:59 PM
re: MPLS Vendors Demo Fast Reroute dreamer:
In the scenario you presented, how does SONET ensure that the failover between ports on both switches happens at the same time? I guess I just don't see how there is any difference between MPLS FRR and SONET APS. Can anyone explain?

best regards,
--Beo

dreamer101 wrote:
"The proper test is to only cut the RX cable. The RX side will detect the failure. It will generate Sonet AIS to the TX side on the same port. Depending on the length of the TX cable, the other side may take a while to receive AIS. Obviously, if the TX cable is a few thousand miles long, it will take a while to switch over. How does MPLS FRR ensure that the port is switchover on both side at about the same time???"
BobbyMax 12/4/2012 | 9:26:00 PM
re: MPLS Vendors Demo Fast Reroute The tests do not tell any thing. The restoration and alternative routing capabilities can be reliabily tested in a large size network. The carriers will never deploy MPLS in the entire network because of the performance problems associated with MPLS.
techstud 12/4/2012 | 9:26:00 PM
re: MPLS Vendors Demo Fast Reroute Tera,

as the previous poster said - mpls hyping serves two objectives for established vendors - 1) keep new entrants busy by making them implement useless mpls stuff that will unlikely be ever completed given their limited resources (eventhough it is unlikely to be ever used on those machines) and 2) sell new "mpls" capable gear by confusing dumb operators by repeatedly telling them its "NEW". Cisco needs to sell tons of mpls add=on cards to old atm switches.

Think of it who will make most money by hyping mpls, and who gets to loose most? small vendors have no choice but get caught in cleverly laid trap.

How many small companies died simply by wasting too much resources on mpls feature development just be able to say to the operators that they have mpls (even though most operators have no plan to use it real networks except for their lab set ups)?
opticalwatcher 12/4/2012 | 9:26:01 PM
re: MPLS Vendors Demo Fast Reroute "When are we going to learn that consensus is always wrong in data comm market??"

To a certain degree who is right or wrong or whether or not MPLS is "dead-end" is irrelevant.

What matters is what customers with money are asking for. And most are asking for MPLS in one form or another, either short term or long term. So the vendors are going to put it into their equipment. Even Wavesmith and Equipe have an MPLS story.


sgan201 12/4/2012 | 9:26:02 PM
re: MPLS Vendors Demo Fast Reroute Hi,
MPLS is very useful of keeping your competitor investing on useless stuff that no one will buy. Then, you can focused on creating something marketable with little or no competition..
Look at situation right now..
RBOC want to buy ATM switch.
There are only 2 next gen ATM switch start up: Wavesmith and Equipe..
Meanwhile everyone waste their money and effort on MPLS...

When are we going to learn that consensus is always wrong in data comm market??

Remember, 15 to 20 years ago everyone is convinced that computer should be inter-connected via PBX. NT, AT&T, Rolm, Siemen and everyone else dump 100 of millions dollars into that and then Ethernet show up and take over the LAN.

Mech4 12/4/2012 | 9:26:02 PM
re: MPLS Vendors Demo Fast Reroute Great test, great results - who cares.....

The 2 things I get from this article is that these 3 cos. did something together, that worked and that Worldcom still gets their thoughts into public forums.
sgan201 12/4/2012 | 9:26:03 PM
re: MPLS Vendors Demo Fast Reroute Hi Tony Li and others,
Speed of light propagation and other stuff do affect in a certain degree. It depends on how the cable is cut. There are two optical cables in each port: TX and RX. The proper test is to only cut the RX cable. The RX side will detect the failure. It will generate Sonet AIS to the TX side on the same port. Depending on the length of the TX cable, the other side may take a while to receive AIS. Obviously, if the TX cable is a few thousand miles long, it will take a while to switch over. How does MPLS FRR ensure that the port is switchover on both side at about the same time???

However, we cannot tell from the article whether the test is done properly by only cutting one cable from the port. Cutting both cables on the same port is a trivial test since both side would have detected the LOS and can switchover immediately...
techstud 12/4/2012 | 9:26:04 PM
re: MPLS Vendors Demo Fast Reroute
MPLS by now is an obsolete junk - who cares, except for certain vendors and individuals who have vested interest to sell their gears or so called expertise (if re-writing old atm papers can be called any thing). The reasons mpls work was started in IETF are no longer applicable.

Only useful tauted application for mpls is carrier bgp/vpns which is the most convoluted solution to a problem for which many other elegant approaches exist. Who needs traffic engineering at 10% utilization of network capacity (in any case there are simple approaches to route around congested links)? DWDM gear is sitting unused if you need more bandwidth and 80% (??) fiber is sitting unlit. So why care these tons of mpls junk?

Whole tag switching at cisco was a reactionary development to deal with the threat from Ispilon's fast packet switching etc. It wasn't based on any revolutionary research. Route look up technology has since that time has matured a lot since 1997 when all this started(software algorithm and hardware loop up engiens - NPS, CAM and specailized implementation). Most mpls drafts including all re-route, crank bank schemes, traffic engineering etc are rehash 80/90s atm/circuit switching papers on similar topics.

Mpls is dead, backward looking technology except that now it comes in new shining packaging and with bold letters "NEW". users beware.

beowulf888 12/4/2012 | 9:26:05 PM
re: MPLS Vendors Demo Fast Reroute lob:
I don't think speed-of-light propagation delays would affect the FRR time. Correct me if I'm wrong, but FRR doesn't provide end-to-end protection for an LSP. How is this any different from SONET APS? FRR and APS both rely on the detection of hardware failures.

Now if this were a test of backup LSPs, I'd have to agree with you. Of course, the details of the testing regimen weren't revealed in the article...

best regards,
--Beo
Tony Li 12/4/2012 | 9:26:06 PM
re: MPLS Vendors Demo Fast Reroute In case you missed the point, the speed of light propagation delays are completely irrelevant in fast reroute because there is NO signalling from the hop before the failure back to the head end of the LSP.

Yes, you could do as well with an IGP _in this topology_, but this result holds independent of the specific topology. The failover time here is simply the amount of processing necessary to install the backup LSP. This time should scale linearly in the number of LSPs that have failed.

Tony
arch_dude 12/4/2012 | 9:26:06 PM
re: MPLS Vendors Demo Fast Reroute lob says:

Now we have absolutely meaningless results (5-10 ms) achieved in a test network with ridiculously simple topology and with no speed-of-light propagation delays.

How illuminating. Bet you could do just as well with plain old I-ISIS or OSPF. Just don't use link keepalives as the sole means of detecting LOS.

Running Dijkstra algorithm for 6 nodes on anything more advanced than Z80 would likely take less than 5 ms :)
----------------
lob completly misses the point. MPLS fast reroute reacts locally. Therefore its reaction time is independent of the network size and of the network topology. In this regard it is just like SONET. By contrast, IGP convergence times depend on the size and topology of the AS.

What this means is that the fast reroute results in a big network will be the same as the fast reroute results on the test network.
achorale 12/4/2012 | 9:26:07 PM
re: MPLS Vendors Demo Fast Reroute This is the maliase of the days - anytime some folks demonstrate some good results, people strip this apart!

I guess ignoring propagation delays, this was a great result. Add in whatever propagation delays you want and once can see that the resultant failover time is impressive.
lob 12/4/2012 | 9:26:07 PM
re: MPLS Vendors Demo Fast Reroute Now we have absolutely meaningless results (5-10 ms) achieved in a test network with ridiculously simple topology and with no speed-of-light propagation delays.

How illuminating. Bet you could do just as well with plain old I-ISIS or OSPF. Just don't use link keepalives as the sole means of detecting LOS.

Running Dijkstra algorithm for 6 nodes on anything more advanced than Z80 would likely take less than 5 ms :)
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE