x
<<   <   Page 2 / 35   >   >>
TelcoDude 12/4/2012 | 10:44:01 PM
re: MPLS Will Outgun ATM, Says Poll Tony dude, Good acedamia argument.

The reality: NO TANGIBLE AUDIENCE . It is well known that the IXC/ILECs have complete seperate groups that manage FR/DSL/ATM network and the IP backbone. Who are going to sell this blissfull OPEX message to?

The evolution to an unified network will not happen anytime soon. Each of these network works is self-suffienct in generating revenues and you know well how bad they want to merge (^& REAL WORLD)

Let me ask me CAN you not reduce opex by having a hybrid backplane with both ATM swithing fabric for ATM cards and a Packet Swithcing fabric for IP cards. Why MPLS?

You think it is cheap to bring nanog dudes upto speed on troubleshooting lsp tunnels?

I don't think MPLS is the answer for convergence. Maybe TE.

VOIP? Don't get me started on irrational exuberance of Chambers "Voice will be free". Do you want your phone on VOIP? Really, I mean Really?
TD
TelcoDude 12/4/2012 | 10:44:01 PM
re: MPLS Will Outgun ATM, Says Poll Probably to you. If you happen to work for an MPLS/ATM startup. which one? Vivace? Mahi? Equipe? Tenor?

Let go BBoy. Look for job with Cisco (:
dhanush 12/4/2012 | 10:44:01 PM
re: MPLS Will Outgun ATM, Says Poll The same ATM switches that were bought 5 years ago
need only minor upgrades, both hardware and software to implement these new services.
---------------------------------------------

1. Are we doing this ? Retrofitting existing ATM boxes ?

2. Is it a good idea to bother about "reusing" ATM hardware ? ( with respect to protocol complexity; label merging for instance )

3. Can this approach of "let go to bed with atm today, tomorrow we can stab it" work ?
why not compete with "atm" directly rather than playing santa claus ?





Tony Li 12/4/2012 | 10:44:01 PM
re: MPLS Will Outgun ATM, Says Poll TE is an easy case: throw bandwidth at the problem.

Big network UUNET managed TE manually and are still doing it.
---------------------------------------------

Throwing bandwidth at the problem is not necessarily the right answer. Lighting fiber is not free. In fact, in many cases, it is very expensive. IP unicast routing protocols are not very good at making use of all available bandwidth. They instead focus on the shortest path, which can easily lead to unnecessary congestion issues.

Big networks feel an amplification of this pain and are precisely the networks that requested the development of traffic engineering capabilities.

Tony
Tony Li 12/4/2012 | 10:44:01 PM
re: MPLS Will Outgun ATM, Says Poll WHY do you want to bring it to the NET layer.

tell me REALLY what MPLS has to offer
---------------------------------------------

The economies of scale come from having only one set of boxes. Having both ATM switches and routers is more complicated and thus increases opex.

To be able to do that, the traffic engineering capabilities present in a connection oriented architecture needed to be absorbed into the routers.

In addition to TE, MPLS also enables the carrier to provide fast reroute and VPN services. Some carriers feel that a fast reroute capability is a necessity for carrying VoIP traffic. They wish to preserve the protection that they had in a SONET architecture. VPN services are another integration and stat-muxing advantage. Carriers can now sell ATM/FR/private line services over the same infrastructure as their IP datagram network.

Tony
Tony Li 12/4/2012 | 10:44:00 PM
re: MPLS Will Outgun ATM, Says Poll Yes, I'm reasonably familiar with the inner workings of APS. It's a fine thing. I agree that lighting fiber is not free and you seem to be agreeing that it is expensive. But if lighting fiber is expensive, then that implies that there is a business case in avoiding lighting fiber unnecessarily. And 'thowing bandwidth' at the problem is exactly what results in an inefficient network.

Let's look at it another way: a typical network (phone or IP) will have a traffic demand matrix. Put input cities on one side and outputs on the other. Some amount of traffic is exchanged between these and we need to provision bandwidth to carry this traffic.

If we look at this matrix in the aggregate, our goal is to provision the least cost network that supports this demand. Now, one way to solve this by throwing bandwidth at it would be to simply create a full mesh of links. But if this means that the links in the network are running at single digit utilization, we're probably spent a whole lot more money than is necessary.

So, the first thing that we want to do is to trim out obvious fat. In an IP network, you can do this until your IGP gives you a localized point of congestion. However, at that point, the IGP will not provide you with the flexibility to precisely place traffic on the various paths of the network. It requires more granularity that is simply unavailable. Thus, we need a more exacting tool that allows us to avoid unnecessary congestion and also avoid wasting resources in our network.

In essence, TE is all about running the network at a higher average utilization while maintaining the exact same supported traffic load.

Note that flow control and queueing are not the same as traffic engineering. They are certainly congestion management techniques and we'd die quickly without them. But they act to suppress the load on the network. That's not the goal. The goal is to support the load. Not avoid it.

Certainly MPLS is not the only solution to TE. ISPs have used both FR and ATM as underlying technologies to provide TE capabilities for their networks. Both have been made to work. But not everyone likes them.

That's the great thing about networking. Everyone can have it their way. There are enough standards that we can all be doing something different.

Tony
Tony Li 12/4/2012 | 10:44:00 PM
re: MPLS Will Outgun ATM, Says Poll
You're assuming that I want to sell. As you noted earlier, I'm a techie. I respond to requests from my customers. These include various ISPs, IXCs and greenfield networks. I have no agenda other than solving my customers' problems in the best technical manner. I have no interest in developing solutions and then finding the problems. There are enough problems that lack solutions.

Yes, I realize that there are currently many different and separate groups in many of the ILECs and I realize that politically, they have little motivation to work with one another. I'm not out to solve internal politics in some major corporation. That's up to them to work out. I'm only there to provide the technical means should they want to drive to more efficient operations. Because separate groups are not really necessary.

Yes, one could reduce capex by having a single God box that spoke both ATM as a switch and acted as a router in the same frame. In fact, I've worked on one of those before. But the reality is that a software solution on a pure router platform performs the same functions that we were asked to deliver in that God box without one tenth of the implementation headaches.

Why MPLS? Because it is the right thing for the network. The hybrid strength of both connectionless and connection oriented architectures present, co-resident, and first class citizens in the same system allows us to use the architecture that is appropriate to deliver the service.

Note one very nice property: if you don't like it, don't use it. No one is forcing you to run MPLS and if you like the role of a traditional ISP, you can be very successful without it.

Do I want my phone on VoIP? Yes, absolutely. Voice will be part of my ISP bill. In the sense that it will not be separately billed from my Internet bandwidth, it can, will and should be free. And I will take free any day of the week, except for my calls to 911. For those, I have my cell. ;-)

Tony
TelcoDude 12/4/2012 | 10:44:00 PM
re: MPLS Will Outgun ATM, Says Poll "Lighting fiber is not free. In fact, in many cases, it is very expensive. "

Do you know what SONET is so reliable? Does Working path and Protect path ring a bell? Dude nothing comes free. Throwing bandwith solves many real telco problems.

Flow control, and queuing are good research topics for solving TE. MPLS is not the only soln. ANd using TE don't push all the other crap VPNs, bla bla .
TD
Holy Grail 12/4/2012 | 10:43:59 PM
re: MPLS Will Outgun ATM, Says Poll
Tony

You admit that you talk from the perspective of an Engineer. GǣEngineer = Someone who builds thingsGǥ

Nothing wrong with that in fact I 100% agree with you!

Problem is that we live in a world where often the most technically elegant designs and solutions turn out to be unsuccessful.

The reasons are nothing to do with Engineering instead they are often political or economic.

If we consider the real world for a moment, it is clear that both politics and economics have some significant bearing on this debate about IP/MPLS/TE/VPN. Further, it is my belief that even though I agree with your engineering analysis regarding the potential GǣvalueGǥ of MPLS, I do feel that these other external factors are tending to conspire against MPLS.

To take some of the points raised on this board for example.

You see the Internet and IP in the context of an evolution of packet switching technologies

GǣIf we can carry over this Darwinian process, we may yet separate the wheat from the chaff.Gǥ

Survival of the fittest then I guess?

Well,in terms of survival of the fittest, how do you determine fittest?

Most elegant network architecture? Or most profitable business? If itGs the latter then you wonGt find many ISPGs surviving and thriving. Why? Because the truth is that whilst the Internet is the most pervasive and most useful packet network (It is also unregulated, see political and economic factors above!), and it is mostly constructed using connectionless IP routers, paradoxically it also happens to be one of the least profitable networks. The irony is that whilst MPLS can enable new and more profitable services, these IP (ISP) only based operators are the ones who can least afford to invest in new edge/core and service management (billing systems) technologies that enable VPNGs. (ItGs not quite as simple as this but you get my point).

Regarding another good point made earlier on this board, the reality that many operators have separate ATM, Frame and IP networks and that this makes it hard to argue for the multi-service convergence business case. Politics and economics here again IGm afraid, both conspiring to some extent against MPLS and the multiservice ATOM plan. In the current economic environment how do you make a business case for delivering the same ATM and F/R services over IP, when you already have an infrastructure written off on the books and installed delivering these services today? The only way to make the business case for actually spending money on MPLS services is where those services are new services and where they generate incremental and high margin service revenues. This is what is so clever about CiscoGs RFC2547, again as an engineer you probably hate 2547? But guess what, it is being deployed! Why? Economics and Politics win again my friend. You know whatGs really hilarious about 2547? Many operators have deployed it on an entirely physically separate IP network, whooho now thatGs what I call economy of scale, way to go Cisco!!

MPLS is a very important evolution within the IP model. It is here to stay and it has adds some nice things.

Tony, please put your engineering hat on and find a useful service that is needed today, is not already being provided by some other network infrastructure that is already paid for, and that requires MPLS to be deployable. Any suggestions?

Cheers



dhanush 12/4/2012 | 10:43:59 PM
re: MPLS Will Outgun ATM, Says Poll How are we going to handle scalable small group multicast ? ( assuming that bandwidth has to efficiently utilised ). Or would we be reverting to supplementary IP-SGM functionality in mpls boxes ?
<<   <   Page 2 / 35   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE