x
<<   <   Page 2 / 5   >   >>
netskeptic 12/4/2012 | 8:26:30 PM
re: Nortel's Empty Room at the Top > As far as Enterprise business, you should be
> aware that the big trend is towards outsourcing,
> where Carrier Managed Services are going to grow
> dramatically.

I am very curious to learn more about this development - can you provide any references ?


> Guess what ? Cisco isn't winning any major
> prizes with Telcos -- especially with their
> focus on Layer 3 which is going to be superceded
> by Optical Ethernet and Layer 2 connectivity at
> the edge.

It does not mean that I disagree with your point, however, I did not see anything on this topic beyond 802.17 efforts, which looks rather like a pipe dream than a serious metro technology. Can you provide any pointers to other development/analysis ?

Again, it does not meant that I disagree, however, it seems to me that Nortel just recently announced its MPLS everywhere (or next generation MPLS) initiative and in my view it does not fit well with your argument.

> What Carrier wants to take on the
> nightmare of provisioning all those meshed
> connections ?

I suppose that unless Cisco (or others) will come up with routing protocols which scale better than BGP routers will be out of the core pretty soon.

Thanks,

Netskeptic
prefer_to_lurk 12/4/2012 | 8:26:28 PM
re: Nortel's Empty Room at the Top netskeptic:

Fair questions -- and I appreciate your input. I probably don't have the hard facts you're after, but I'll at least try and explain what led me to my opinions...

1) Carrier Managed Services
- Pretty much an industry trend, backed by independent market research -- but no, I don't have a link to any publically available reports. I've also had direct experience with numerous large Enterprise customers wanting a full Carrier Managed Solution. It does strike me as a logical extension of the Hosting Center, ASP and SAN type scenarios.

2) Regarding RPR/802.17
- Working for Nortel, I'm biased here, but the technology that we've submitted as a candidate for the eventual 802.17 standard has been working in live networks for over a year. In fact, OPTera Packet Edge technology goes beyond the core features outlined for RPR (spatial reuse, L2 protection, etc) and can provide a true Ethernet UNI as well as a connectionless Layer 2 architecture. There is a good deal of information on this solution on the nortelnetworks.com website.

3) MPLS
- I'm not well versed in the complexities of MPLS, but I believe it can be complementary to the Optical Ethernet push. Ethernet can be used in the Access network to transport and aggregate traffic at a POP or Switching site. From there it could enter the MPLS core through a core switch or router. I'd like to understand this better, so educate me if I'm way off base here.

Thanks for your reply ! I appreciate the dialog and the chance to learn something new (as opposed to the all too typical "Company X sucks! Company Y rules!" posts).

ptl

lordcirrus 12/4/2012 | 8:26:28 PM
re: Nortel's Empty Room at the Top > What Carrier wants to take on the
> nightmare of provisioning all those meshed
> connections ?

My experience is limited to only the OM3000, but I really don't see what the big deal is here...

It might take like an hour the very first time, during installation/commisioning, but once the box is up and running, it should be easy. If anything, this is indication that the telcos aren't managing their networks efficiently, rather than anything else. And the tools for single-vendor multi-product, 5-click end-to-end connection building is coming...
netskeptic 12/4/2012 | 8:26:24 PM
re: Nortel's Empty Room at the Top prefer_to_lurk
> Fair questions --

Thanks, I will look at all pointers provided.

I would like to ask another question wrt 802.17 before I got to notel site: do you have something like ABR (which I deem necessary for any successful L2 technology) or it is still
strictly CBR + UBR ?

Thanks,

Netskeptic

wildcard 12/4/2012 | 8:26:23 PM
re: Nortel's Empty Room at the Top Skeptical indeed the 802.17 address the control and shaping control into the connectionless Optera world. The whole issue of the SLA has to do with mixed service offerings from my desktop/laptop/portal/ what ever they call it today not the WAN ingress. The Nortel and Cisco differences need to be put to rest. It could be John Chambers that takes the helm at Nortel in 2002. Nortel has just as good a shot at the premise push through the cornerstone family, as the Optera dream fades in a prolonged engagement with Cisco. All you designers and would be network geniuses please try and remember the last/first mile and who pays for your habits.

Real food for real people virtually speaking.
prefer_to_lurk 12/4/2012 | 8:26:23 PM
re: Nortel's Empty Room at the Top netskeptic:

ATM service categories don't conceptually map into the connectionless OE model. While outside of the proposed 802.17 standards, Nortel has implemented traffic shaping and rate enforcement directly on the OPTera Packet Edge cards (as opposed to Cisco, which requires another box sitting in front of the 15454).

Through this "Ethernet UNI" a customer can be given both a guaranteed data rate and a maximum burst rate, while the Ethernet ports can also send PAUSE frames during ring congestion. Overall, the capabilities are quite similar to the intent of ABR.

Do these capabilities address your concerns, or is the technology missing something ?

ptl
gladysnight 12/4/2012 | 8:26:22 PM
re: Nortel's Empty Room at the Top "Yes, I am just wondering is it going to be good enough for carriers in the long term?"- netskeptic

I personally doubt it.

I believe that bulk carriage of differentiated packet traffic is like the e-commerce advertising mantra "mass customisation" - i.e. inherently expensive and possibly futile and/or impossible.

There are good reasons why market segmentation happens, why it "works" in planning and deploying service offerings, and why most (all?) businesses do it.

Something about the law of diminishing returns . . .
netskeptic 12/4/2012 | 8:26:22 PM
re: Nortel's Empty Room at the Top > netskeptic:

> ATM service categories don't conceptually map
> into the connectionless OE model.

I understnad that, however, (as a former Digital hand) I am a true beleiver that effective network mechanism should include a possibilty to effectively re-use unallocated and unused bandwidth with following characteristics: per stream fairness, no head blocking and no packet loss, and I know this is quite costly proposition.

The alternative in my mind is to maintain excessive network capacity at all times, which may prove to be a cheaper deal, e.g. the current Internet uses this approach with a great success anyway.

Funny thing is that if this is OK long term policy, then we ain't need no stinky RSVPs, and diffserv will do just fine with all our networking needs.

I am wondering are there anybody else sharing my opinion ?

> While outside
> of the proposed 802.17 standards, Nortel has
> implemented traffic shaping and rate
> enforcement directly on the OPTera Packet Edge
> cards (as opposed to Cisco, which requires
> another box sitting in front of the 15454).
> Through this "Ethernet UNI" a customer can be
> given both a guaranteed data rate and a maximum
> burst rate,

This is reserved bandwith and it simple, clear and
straighforward.

> while the Ethernet ports can also send PAUSE
> frames during ring congestion.

So, you opted for bulk fairness, head blocking and packet loss, compensated by simplicity, big buffers and bandwidth glut.

> Overall, the
> capabilities are quite similar to the intent of > ABR.

I would say that these are quite below original intent of ABR.
From another side ABR did never happen anyway :):).

> Do these capabilities address your concerns, or
> is the technology missing something ?

Yes, I am just wondering is it going to be good enough for carriers in the long term ?

ptl

Thanks,

Netskeptic
prefer_to_lurk 12/4/2012 | 8:26:20 PM
re: Nortel's Empty Room at the Top netskeptic:

I'm not trying to be argumentative, just trying to keep up with you...

Reuse of unallocated bandwidth
- I understand this was a (the ?) key attribute of ABR. However, can you not somewhat compensate for this through network engineering ? By choosing appropriate ring bandwidth(OC-12, OC-48, OC-192), RPR bandwidth (STS-1, STS-3c, etc) and number of sites -- it should be possible to maintain a high utilization rate of network resources. Not as flexible or convenient as ABR, but probably a whole lot cheaper.

Reserved bandwidth
- Since OE is not connection-oriented, you still have the option of oversubscription even for "Guaranteed Bandwidth". Since there is no concept of a circuit, you don't really ever need to have assigned but unused bandwidth. Carriers may need to play a careful statistics game to avoid violating SLAs, but there shouldn't be excessive unused bandwidth laying around.

Bulk Fairness, Head Blocking and Packet Loss
- OPTera Packet Edge has several mechanisms in place for flow control and class of service support. This includes token buckets, ring insertion and ring fairness algorithms. I apologize, but instead of giving out too much proprietary information, I'll steer you towards the following public document for more details:

http://www.nortelnetworks.com/...

Not as robust as ATM, but again I believe the OPTera Packet Edge technology makes out very well in a price/complexity versus performance comparison.

I'm probably coming at this more from a non real-time data perspective, and therefore question the need for complex ATM-type traffic management. On the other hand, I have to admit that I don't fully understand the implications of trying to carry real-time traffic over a wide-area Ethernet.

But even for voice or video traffic, is it not up to the (Layer 3) end devices to develop the necessary signalling techniques and traffic management schemes ?

One other data point: measurements from live OPTera Packet Edge customers have shown an IMPROVEMENT in latency over the legacy ATM networks.

Just to reiterate, I'm trying to better understand the potential issues you have identified and consider this an educational opportunity. Thanks for your continued comments.

ptl




netskeptic 12/4/2012 | 8:26:20 PM
re: Nortel's Empty Room at the Top gladysnight
Number: 18
Subject: Re: the race to chapter 11 is on!
Date: 5/13/2001 8:46:39 PM

"Yes, I am just wondering is it going to be good enough for carriers in the long term?"-
netskeptic

> I believe that bulk carriage of differentiated
> packet traffic is like the e-commerce
> advertising mantra "mass customisation" - i.e.
> inherently expensive and possibly futile and/or
> impossible.

It does not mean that I disagree, however, I am not sure that I got it rigth.

I suppose, you point is that carriers will allocate badnwidth in the form of independent networks/pipes on a per-customer basis. If customer is not using allocated bandwidth then so be it - is it correct ?

Thanks,

Netskeptic

<<   <   Page 2 / 5   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE