x
<<   <   Page 2 / 14   >   >>
sgan201 12/5/2012 | 3:39:35 AM
re: BT Rethinks 21CN Core Strategy 1) There is no big mesh here. The VC/tunnel form a Ethernet connection to the customer. It is a Ethernet point to point virtual leased line. At least, that is what I can gather from this article.

2) At least in ATM, you have very well defined traffic management capability. Now, we lost that. Maybe it is okay since we have so much bandwidth at the core.

Dreamer
jhawkins 12/5/2012 | 3:39:34 AM
re: BT Rethinks 21CN Core Strategy Well thanks! I think... (wonder if my boss reads this?)

Not sure your confidence in our marketing prowess is well-placed though. BT guys are definitely WAY smarter than that.

BTW, don't confuse this with VLAN cross-connect. That's another criter altogether.

It may help to think about this differently. Don't think in terms of a challenge to MPLS per se. MPLS services (2547, PWs, VPNs, etc) are well thought-out, valuable parts of the NG network offer. We'd just like to run them over native Ethernet, that's all. Save the complexity on the control/signaling functions, and let Ethernet be Ethernet (ok, within reason).

That should leave more $$ to spend on something else. Who knows? Maybe we sharp marketing guys can figure out how to get some hapless customer hooked on whatever THAT is! Now that IS good marketing. :-)

john
yhza 12/5/2012 | 3:39:33 AM
re: BT Rethinks 21CN Core Strategy Fixed Circuits, P2P (no any-2-any), no auto-oversubscription, flat topology, propriety (for now), SDH-like-framing, etc

Obviously, an old dog can't learn a new trick. This is nothing but SDH in disguise.

gigeguy 12/5/2012 | 3:39:33 AM
re: BT Rethinks 21CN Core Strategy Hey, I'm the biggest suppporter of Ethernet there is, but I'm afraid that PBT requires quite a bit of scepticism.

I view PBT first and foremost as a marketing last gasp from Nortel given their complete product failure in routing and MPLS ... they're trying to stay in the game via mindshare, marketing, and a bit of competence that they have left in bridging.

It's been estimated that only perhaps 10% of existing Ethernet chipsets could support PBT, and there's not even a project yet at IEEE to standardize it, so it's just proprietary Nortel technology right now.

BT got roped in because they think they can bring down their capex costs because "Ethernet costs less". Well, many times you get what you pay for. You have to pay for carrier-class HW and SW resiliency, whether it's in a router or a switch, if you want to be able to offer your customers an uptime SLA.

You also have to remember that PBT turns Ethernet switches into label switches. You still need a control plane, and the PBT supporters are working in the IETF GELS working group to adopt the MPLS control plane for PBT.

Well, where's the complexity in MPLS? Not in the forwarding plane - that's just swapping labels. It's in the control plane signaling. So why is PBT going to be any less complex than MPLS at the end of the day?

<end mode="" of="" rant="">
</end>
SolitonWave 12/5/2012 | 3:39:33 AM
re: BT Rethinks 21CN Core Strategy I agree that Ethernet will rule. You're assuming ethernet and MPLS are exclusive options. I think they can live together (complementing each other, without PBT).
I did not say PBT it is not real. I just argue it's not much more than a clever combination of "fairly straightforward technical solutions"+"pretty good marketing".

PS: Are BT "serious investigators" the only clever guys around? I'm asking because many other "serious investigators" are backing other options... If you read the press it seems PBT is the winning horse in race track. The fact is: it is not adopted by 802.1 (yet). It is not adopted by MEF (yet). And I say "yet" because with the right Marketing anything can happen.
yarn 12/5/2012 | 3:39:32 AM
re: BT Rethinks 21CN Core Strategy Indeed it makes you wonder. If PBT leverages existing Ethernet technology (and even removes features), then why wasn't it invented 5-10 years ago? Was it so simple it was overlooked, or were all hopes still set on MPLS?
The cost of a technology is less determined by its inherent complexity than by the level and scale of industry adoption. How is PBT making a transition from a niche to a mass market? Because BT says so or only when it indeed obsoletes MPLS?
davallan 12/5/2012 | 3:39:32 AM
re: BT Rethinks 21CN Core Strategy Hi Gigeguy

I guess I do not hear from the 90% of silicon vendors you talk to. Given the required changes to ethernet behavior are trivial and publically documented (we are discussing modifying a few clauses in 802.1Q here), it is hardly proprietary in teh sense of "incorporating trade secrets."

You can consider the MPLS dataplane simple if you forget MPLS by definition is grafted onto existing IP router interfaces. a.k.a. The toothpaste T-MPLS proponents are trying to stuff back in the tube...

The other aspect of MPLS which adds complexity is arbitrary label stacking at any point. When you work through the operational and implementation impacts of that vs. keeping a given Ethernet domain flat IMO there is a structural difference that is maintainable even when the addition of a control plane is considered.

cheers
gigeguy 12/5/2012 | 3:39:32 AM
re: BT Rethinks 21CN Core Strategy davallan,

Could you explain how one can scale networks, especially between providers, with a flat address space and no means of hierarchy other than pushing and popping MAC headers at 20 bytes per header, vs. 4 bytes per label?

And even with just a single level, there's 4 times the per-packet overhead, which can be significant when near 50% of the packets being transported are just TCP ACKs. Small VOIP packets make the overhead even worse - and we used to complain about the ATM cell tax!!!!

And I didn't hear any commment on the fallacy that PBT is going to magically save carriers on their CAPEX?

Looking forward to hearing back ...
davallan 12/5/2012 | 3:39:31 AM
re: BT Rethinks 21CN Core Strategy HI Gigaguy:

Simplest analogy is that PBT over an Ethernet link looks like PHP'd MPLS. 4 fewer bytes as there is no label, not 20 more.

And we're only removing the Ethernet header at the tunnel end points, not removing and reinserting at every hop. BTW Same analogy can be made for POS L1 overhead...

Finally good specification leads to commodization, somthing IMO the IEEE has done well...

later
D
Mark Seery 12/5/2012 | 3:39:29 AM
re: BT Rethinks 21CN Core Strategy A question and then some comments.....

As the ability to reuse the existing Ethernet technology base is central to the argument for PBT (at this point in time) it would be interesting to hear the argument behind the assertion that only 10% of chips can support PBT (is this related to MAC-in-MAC, OAM, QoS,...??).

Some comments.....

There are a million angles on this issue, but to hit a few of them raised on the thread:

-it is not inaccurate, IMO, to say BT was roped in to this. They have done extensive theoretical research in to network architectures (for many decades), were looking for something else other than MPLS (well at least a pocket within BT), and played a large role in pushing the modeling of Ethernet-based architectures (in a telecom ecosystem sense).

-on why was this not created previously.....because the carrier part of the industry has only in the last few years started to do modeling of Ethernet networks in any serious way. additionally, until it became a slam dunk that the future was Ethernet, then the focus of networking necessarily remained how to label switch IP (and later other protocols) over multiple subnetworks (and hence MPLS).

-I agree that PBT is a replacement for SDH, but I don't agree it is SDH in disguise. PBT is both better and worse than SDH (IMO).

-If someone has an opinion that MPLS is not complex in the data plane, they are of course entitled to that, however I don't believe that opinion is shared by people who do not like MPLS.

-Lower capex can be achieved by both the inherent cost of a technology (allowing for telco hardening costs as stated) as well as the presence of more supplier competition.

-you don't have to make a forwarding decision based on the addresses in IP header, to make use of information in the IP header.

-the three great historical domains in telecom networks were switching, transport, and access (an the INs and OSSs supporting them). the question on the table is what does the future look like.

-all technologies break down problems in to areas, domains, layers etc to achive scaling. BT is concerned about scaling both from a cost and technical perspective, and PBT is not the only technology they are exploring from this perspective. You don't necessarily have to solve the entire scaling problem in one domain.

-it is important to at least appreciate the non-technical issues of all this, like who is and is not lined up in support of MPLS, the role of carriers in driving standards, etc.....but that would take a long dissetation in and of itself.
<<   <   Page 2 / 14   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE