x
Mark Seery 12/5/2012 | 3:01:19 PM
re: PBT vs MPLS: Round VII 1. On modern day routers

"Here are the next steps for PBT:

- a signaling protocol
- a topology distribution protocol
- an OAM protocol
- a link resiliency protocol
- QoS definitions

Guys, how is that NOT a modern day router? The logic defies me."

Well the above sounds a great deal like an ATM switch to me. Would you call an ATM switch a modern day router? Or would you call it a subset of a modern day IP/MPLS router in the form of MPLS? For example, what does any of the above suggest, if anything, about the desireability to for example virtualize NAT to provide a network-based IPVPN service? What else do IP routers do that have little to do with the above?

2. On routers as TDM switches

In so far as there is a difference between scaling millions of dynamic/short hold time subscriber to subscriber connections versus maintaining a long-hold connectivity matrix between only thousands or tens of thousands of locations an analogy can be drawn between the complication and justifiable risks of a voice switch control plane and IP routing, and correspondingly the potentially different mission of a transport-oriented control and/or management plane. However, my opinion is that it is a stretch to go beyond that. I would view IMS and/or competing web technologies as being much more the center of service features that are analagous to a TDM voice switch than what a router provides. In so far as there is a limited number of suppliers for a given technology then there is perhaps an additional analogy - but is this not what the market is in a sense addressing by creating multiple competing and fully/partially substitutiable (economically and technically) products and technologies for a given function?

3. On the "core"

In both this thread and a related thread, the term "core" is used like some kind of precise definition of something. Kind of reminds me of a statement such as "round up the following number: 1.444657858". Kind of meaningless unless you are making a significant and potentially error prone assumption.

Only those that believe there will only be one transfer protocol and layer could possibly use the term "core" in this way, from my perspective. It is like a closed model of networking being self-referential. Not just from a technical perspective but from an economic one as well. It is like suggesting a market needs one product at one price and all other ideas be damned. I suspect this market, like many others, will have the last say on this subject.

Same goes for "edge". Edge of what.

4. On the cost difference between IP/MPLS routing and PBB/PBB-TE

Many equipment vendors have multiple line cards and/or multiple products addressing different price points. If there are not any real cost differences between these offerings, then I see a potential problem for equipment vendors down the road in multiple dimensions as they try to explain their approach and pricing structure to operators. Either it does or it does not impact the cost of a product to provide large scale BGP and IPVPNs (a potentially separate problem domain from Ethernet transport/services). If it does not, then operators should be asking "why all these different product variations?"
litereading 12/5/2012 | 3:01:21 PM
re: PBT vs MPLS: Round VII Ray, you might have misunderstood my point comparing routers and TDM switches - I wasn't implying that routers would go away because of PBT. What I was trying to convey was Telco's became so reliant upon TDM Switches that the few equipment vendors who made them were able to exert control over the service provider ability to offer services to their end customers and were able to sell the equipment at high margins. Eventually, telcos found their businesses in jeopardy by alternative technologies. Routers in a way have started to exert the same effect - controlled by a few vendors charging high margins exerting a strong influence on the services these service providers can provide.
CoreRouterBuilder 12/5/2012 | 3:01:23 PM
re: PBT vs MPLS: Round VII Something like this will do the trick:

draft-martini-pwe3-802.1ah-pw-00.txt

In PBT they plan to use TCP based LDP (not RSVP TE) or manual provisioning for setting this up. People talking about TE; In MPLS TE, connection oriented ip is done by using extensions to OSPF TE and IS-IS TE. GÇôsource routing

I wander how they would want to do automatic traffic engineering in PBT end to end; I just do not see all the components to make this happened. Multiple Spanning Trees (MSPT)? well

I am not against innovation but it seems a lot of work needs to be done here for something like this to work
digits 12/5/2012 | 3:01:24 PM
re: PBT vs MPLS: Round VII "Could it be that core routers have become the equivalent of yester-years TDM Digital Switch?"

I think the debate is in danger of spilling into fantasy territory. It's clear there is interest in PBT, but the router vendors aren't staring oblivion in the face.

Here's the thing I have noticed -- in public, I haven't heard anyone suggest that PBT could usurp MPLS from the core network, or even replace it entirely in a metro network. YET - this seems to be a scenario that people, mostly from the anti-PBT camp, talk about.

I suspect that some pro-PBT players might be overstating PBT's appeal and potential, and that's something that could work against the technology's acceptance.

It's early days still, but I get the feeling this topic has legs.
CoreRouterBuilder 12/5/2012 | 3:01:24 PM
re: PBT vs MPLS: Round VII You are absolutely right. (Nortel has a good track record in routing: Wefleet; CR LDP push in IETF).

Nortel should go to ITU to make PBT as a new standard
schlettie 12/5/2012 | 3:01:26 PM
re: PBT vs MPLS: Round VII The next steps for PBT (signaling/routing protocol, OAM, QoS) are already being defined in IEEE/IETF/MEF.

The manufacturing cost of these PBT boxes will be only marginally less than the cost of a carrier ethernet router: both need the same fabric, redundant CPUs, fans, cooling, etc. The only cost difference might be due to the lack of line rate IP forwarding, which isn't the rocket science it was ten years ago.

The main difference between a PBT box and a router will be sales price: operators want to built out their transport networks with low-software content boxes that they can pay 30 points margin on, rather than the 65 points they have to pay Cisco & Juniper. Of course, once Nortel & co succeed in bloating PBT into a poor-man's MPLS, the business case for PBT (low NRE for a low margin product) will be broken.

And what operator in their right mind would trust Nortel to deliver a scalable routing stack?
rjmcmahon 12/5/2012 | 3:01:28 PM
re: PBT vs MPLS: Round VII Anyway, what's with this whole notion that the edge is cheap? Nowadays, all the heavy lifting is done at the edge,

If I had to guess people are looking at it from a unit volume perspective and not an NRE perspective. Also, in the current model the service providers buying edge equipment have the stronger position when negotiating price which means edge vendors will always run on the verge of bankruptcy and continuously be pressured to reduce gross margins (assuming arms length business transactions are the forces at hand.) Service providers are trying to repay capex with net income from an overly supplied, multichannel, entertainment model. Enterprise markets are much different because GDP improvements by the corporate staff rationalize the capital spending and the spending on technology churn. I'd short edge vendors until something drastic happens to the operating models.
desiEngineer 12/5/2012 | 3:01:28 PM
re: PBT vs MPLS: Round VII Craig,

"Regarding the core ... Mostly, these non-routers are headed for the edge. I wasn't at Geneva, but all the PBT talk I've heard lately has edge-focused, with the concession that Cisco/Juniper routers still rule the core."

But what is the box in the middle of the cloud? It's still a PBT box - unless PBT is so marginalized that it is only the edge node.

PBT is supposed to be a cheap ethernet switch, with simple SDH-like provisioning. You don't SDH-provision just one box at each end. So what's the point of putting a huge MPLS-based network in the middle?

PBT can only win (i.e., be a profitable venture for a vendor) if there is a sizeable metro cloud of PBT nodes.

All the MPLS-based carrier ethernet edge vendors play the same game with the core nodes, i.e., MPLS features, IGP, etc.

Anyway, what's with this whole notion that the edge is cheap? Nowadays, all the heavy lifting is done at the edge, and the core node is BGP-free, PIM-free, with just aggregate QoS, and is pretty much just a brute. Not to say it is easy to build or cheap - just that the edge is where the VPNs, the per flow QoS, the policy enforcement, etc., is done.

-desi
desiEngineer 12/5/2012 | 3:01:28 PM
re: PBT vs MPLS: Round VII Craig,

"You've got a point, considering Alcatel even calls the 7450 a "router" sometimes. But I think "not a router" refers to the absense of BGP or OSPF or whatnot."

Well, the 7450 has OSPF, so is it a router? It has MPLS, so is it a router? I'd hesitate to call a box that can do label-switching a router unless it could actually signal the labels, and use an IGP.

A better definition than "can it do OSPF" would be whether it can really do OSPF or does it play one on TV, viz. is it a full-function OSPF, does it scale, can it do lots of areas with lots of LSAs, do you have to stick in a new blade for each additional 100 LSAs, etc.

Likewise with MPLS: can it do RSVP & LDP? How many LSPs or FECs? Can it scale, can it FRR a reasonable number of LSPs in sub-50 ms, etc.

That's how you tell a firewall which does a little OSPF from even a low-end router. Or an edge aggregation MTU from a real router, regardless of whether it can do a little bit of MPLS.

-desi
Pete Baldwin 12/5/2012 | 3:01:31 PM
re: PBT vs MPLS: Round VII You've got a point, considering Alcatel even calls the 7450 a "router" sometimes. But I think "not a router" refers to the absense of BGP or OSPF or whatnot.

(Basically, I think it's code for "Not built by Cisco or Juniper.")

Regarding the core ... Mostly, these non-routers are headed for the edge. I wasn't at Geneva, but all the PBT talk I've heard lately has edge-focused, with the concession that Cisco/Juniper routers still rule the core.

Related: Would a pure layer 2 Juniper box still cost more than a Ethernet services box? I'm thinking it would...
litereading 12/5/2012 | 3:01:32 PM
re: PBT vs MPLS: Round VII Could it be that core routers have become the equivalent of yester-years TDM Digital Switch?

digits 12/5/2012 | 3:01:33 PM
re: PBT vs MPLS: Round VII Touch+¬...

Router envy -- I feel that could become standard industry parlance.
desiEngineer 12/5/2012 | 3:01:34 PM
re: PBT vs MPLS: Round VII No, you have to get away from the conceit that Nortel could be involved in anything useful.

OK, so don't call it a router. Call it an ethernet services switch (oh wait, Alcatel uses that name). Here are the next steps for PBT:

- a signaling protocol
- a topology distribution protocol
- an OAM protocol
- a link resiliency protocol
- QoS definitions

Guys, how is that NOT a modern day router? The logic defies me.

It's a box of the same genre as routers today that have ethernet capabilities coupled with MPLS: 7750/7450, M320/MX 960, 7609.

-desi
Marc Rapoport 12/5/2012 | 3:44:18 PM
re: PBT vs MPLS: Round VII Desi,
I think the debate is not happening in the Edge, which is already dominated by MPLS PEs which are features rich (VPN, QoS, ...) and not cheap, but at the Metro, which is the backhaul from the access to the edge.
That's where PBT is claiming that a dumb/cheap ethernet pt2pt cct can do the trick in this layer (which i personally disagree)
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE