- Test - lists.nanog.org

Recommended NON-VIP based 7500 swtching paths
by Brandon Applegate 28 May '99

28 May '99

Greetings, I was looking to get some feedback as to what people are using in production on 75XX routers on standard line cards (HIP for example) for switching paths. I read one time on Cisco's site that on a 7500 non-vip router, you can see up to 150k PPS using plain fast-switching, and up to 275k by using optimum. Could someone be kind enough to elaborate on the differences between fast/optimum switching other than the addition of 'optimum' on the command line ?? I dont remeber seeing too much detailed info on the differences of these earlier switching paths on CCO. Feel free to reply directly via email/off-list. Thanks in advance. -------------------------------------------------- Brandon Applegate, CCNA : Network Administrator http://www.one.net : brandon(a)one.net --------------------------------------------------

2 1

Re: OSPF multi-level hierarchy: Necessary at all?
by Alex Zinin 28 May '99

28 May '99

At 15:11 28.05.99 +0400, Alex P. Rudnev wrote: >Btw. Flat network... > >Routers now have - 128 or 256 MB RAM, 300 - 400 Mhz CPU. I guess you can >built flat betwork with 5,000 routers withouth hard problems. Mmm...I wouldn't be so sure, actually. Recall we need to refresh each LSA every 30 minutes not to let some bugs drive our network crazy. Also the more routers you have in the flat network the greater is the size of LSDB (or whatever table is used) and, since all routers see all other routers and links, the more CPU time we need to spent calculating our SPFs, maintaining the RT, etc... > >Through it's not the question. There is _already_ 2 levels; you can use >multi-zone OSPF (independent OSPF networks connected by _redistribute_). >The question was _is 2-level hierarchy enougph__? Externals in OSPF can also become a scalability issue---Type5 LSAs have domain-wide flooding scope....so, unless you want your areas to be stub, you have all your externals crossing all areas and installed into LSDBs of all routers, making them spend more time on SPF, while it is understood that the lower the level of hierarchy a router is in, the less granularity of external routing info it needs. Alex. ------------------------------------------------------------------ Alex D. Zinin, Consultant CCSI #98966 CCIE #4015 AMT Group / ISL Cisco Systems Gold Certified Partner http://www.amt.ru irc: //EFNET/#cisco, //irc.msn.com/#NetCisco [Ustas]

1 0

Re: OSPF multi-level hierarchy: Necessary at all?
by Alex Zinin 28 May '99

28 May '99

At 15:33 27.05.99 -0700, Vadim Antonov wrote: >Tony Li <tony1(a)home.net> wrote: > >>I suspect that the main driver is not the amount of routing information >>in the gross sense, but the scalability of the protocol as the number of >>nodes increases. > >There's a better solution: decrease the number of nodes by replacing >clusters with bigger boxes. This has an additional advantage of reducing >number of hops (and, consequently, latency variance). "Have more bigger boxes rather than less smaller ones"-approach is not for everybody and not for every case. If you have clusters sitting in one room, powered from the same source, sharing the same ceiling that can fall, running the same version of soft, using the same config., etc., than yes it's ok, because they will more likely crash at the same moment. Otherwise, I'd think first. Also, even if you do use a large box, you probably don't wanna know all the details about it's connections at some level of your network. > >PS. Using DUAL or DASM instead of SPF helps, too -- these algorithms tend > to eliminate updates which "do not matter" unlike SPF-based algorithms > which have to inform everyone about local topology changes. In SPF-based protocols we have areas for this purpose---we do not propogate topology information across the area boundaries. Also, magically we don't have SIAs in SPF even when there's a lot of traffic :) ------------------------------------------------------------------ Alex D. Zinin, Consultant CCSI #98966 CCIE #4015 AMT Group / ISL Cisco Systems Gold Certified Partner http://www.amt.ru irc: //EFNET/#cisco, //irc.msn.com/#NetCisco [Ustas]

1 0

Re: OSPF multi-level hierarchy: Necessary at all?
by Sean Donelan 28 May '99

28 May '99

avg(a)kotovnik.COM (Vadim Antonov) writes: >Well, actually it is not that bad. The biggest number of locations is >probably found in AT&T phone network - 250 or so. Sprint is in few >dozen. The existing IGPs are quite happy with that kind of complexity, >so if you belong to the "one-router-per-POP" school of thought the >IGP complexity is a non-issue. Well, the phone system is already hierarchial. The top-level is pretty small, it is the second and third levels which are monsters. There are about 100,000 NXX's in the country. California has the largest state with about 12,000. The Los Angeles LATA is the biggest at about 5,000. Within PacBell there are about 800 CLLI locations for the Los Angeles LATA, not including all the other CLECs who may have locations in LA. Getting from the relatively few IXC access tandems to those 800 locations is the trick. I may be doing something wrong, but I've found OSPF gets a bit cranky with far less than 800 routers and 5,000 routes in an area. -- Sean Donelan, Data Research Associates, Inc, St. Louis, MO Affiliation given for identification not representation

2 1

RE: OSPF multi-level hierarch: side question
by Dan Rabb 28 May '99

28 May '99

Routers will inevitably fail. The question becomes how much exposure do you want when it does? Placing large amounts of customers on a single box is more economical, and is long as you have an uplink to your network with enough bandwidth to support them it's not a problem, but how many customers do you want down when a single router fails? This is obviously more of a political question that an operational one. Dan Rabb > -----Original Message----- > From: Steve Meuse [mailto:smeuse@bbnplanet.com] > Sent: 28 May, 1999 1:59 AM > To: Vadim Antonov > Cc: nanog(a)merit.edu > Subject: Re: OSPF multi-level hierarch: side question > > > > At 03:33 PM 05/27/1999 -0700, Vadim Antonov wrote: > > > >Tony Li <tony1(a)home.net> wrote: > > > >>I suspect that the main driver is not the amount of routing > information > >>in the gross sense, but the scalability of the protocol as > the number of > >>nodes increases. > > > >There's a better solution: decrease the number of nodes by replacing > >clusters with bigger boxes. This has an additional > advantage of reducing > >number of hops (and, consequently, latency variance). > > > >K.I.S.S. rulez :) > > > >--vadim > > Side question: > > At what point do we stop aggregating customers onto a single box? The > technology exists now to have hundreds if not thousands of > customers on a > signle box, but, Do we want that many? > > -Steve > > >

1 0

Re: OSPF multi-level hierarchy: Necessary at all?
by Vadim Antonov 28 May '99

28 May '99

Tony Li <tony1(a)home.net> wrote: >> There's a better solution: decrease the number of nodes by replacing >> clusters with bigger boxes. This has an additional advantage of reducing >> number of hops (and, consequently, latency variance). >Certainly that helps, but as we move forward you still end up with lots of >boxes in the network simply because you're in many different geographic >locations. Well, actually it is not that bad. The biggest number of locations is probably found in AT&T phone network - 250 or so. Sprint is in few dozen. The existing IGPs are quite happy with that kind of complexity, so if you belong to the "one-router-per-POP" school of thought the IGP complexity is a non-issue. --vadim

1 0

Re: FIBER: CA, KS, LA, TX fiber cuts
by Sean Donelan 27 May '99

27 May '99

Fiber cut digest report for May 27, 1999 Hallsville, TX 9:30am CDT ???, LA 10:00am CDT Fairfax County, KS 2:35pm CDT Long Beach, CA 5:00pm CDT -- Sean Donelan, Data Research Associates, Inc, St. Louis, MO Affiliation given for identification not representation

1 0

FIBER: KS, LA, TX fiber cuts
by Sean Donelan 27 May '99

27 May '99

Fiber cut digest report for May 27, 1999 Hallsville, TX 9:30am CDT ???, LA 10:00am CDT Fairfax County, KS 2:35pm CDT -- Sean Donelan, Data Research Associates, Inc, St. Louis, MO Affiliation given for identification not representation

2 1

RE: OSPF multi-level hierarchy: Necessary at all?
by Jessica Yu 27 May '99

27 May '99

The main purpose of doing hierarchical IGP (OSPF or ISIS), in my view, is to scale IGP routing and achieve fault isolation in large scaled networks. Most of the large networks exit today (especially those using IS-IS as IGP) are still having one flat IGP routing area since IS-IS simply does not have adequate 2-level support. The protocol issues are being addressed in the IETF. I can see a move to 2-level hierarchy in the near future but do not see more than 2 levels would be needed soon. In addition, some large ISPs are moving away from ATM overlay model to MPLS, this will reduce IS-IS/OSPF adjacency greatly (no more full mesh topology from IGP's point of view) thus alleviate IGP scaling issue. Further, the complexity of protocol and operation management will increase with the number of levels added to the hierarchy. In sum, the use of more than 2-level of IGP hierarchy is less likely, at least not in the near future. The following Internet-draft include some discussions of the topic in discussing routing scaling issues in general. Title : Scalable Routing Design Principles Author(s) : J. Yu Filename : draft-yu-routing-scaling-00.txt Pages : 24 Date : 15-Apr-99 A URL for this Internet-Draft is: http://www.ietf.org/internet-drafts/draft-yu-routing-scaling-00.txt --Jessica - ------- Forwarded Message Return-Path: owner-nanog-outgoing(a)merit.edu Received: from mail.nyp.ans.net (mail.nyp.ans.net [147.225.190.25]) by cannes.aa.ans.net (8.8.5/8.8.7) with ESMTP id LAA08768 for <jyy(a)cannes.aa.ans.net>; Thu, 27 May 1999 11:40:17 -0400 (EDT) Received: from segue.merit.edu (segue.merit.edu [198.108.1.41]) by mail.nyp.ans.net (8.8.7/8.8.7) with ESMTP id LAA29164; Thu, 27 May 1999 11:40:11 -0400 (EDT) Received: by segue.merit.edu (Postfix) id 7BB1044439; Thu, 27 May 1999 11:38:43 -0400 (EDT) Received: by segue.merit.edu (Postfix, from userid 56) id 5371344445; Thu, 27 May 1999 11:38:43 -0400 (EDT) Received: from amtsun.amt.ru (amtsun.amt.ru [212.111.64.19]) by segue.merit.edu (Postfix) with ESMTP id 881D144439 for <nanog(a)merit.edu>; Thu, 27 May 1999 11:38:35 -0400 (EDT) Received: from zinin (amtsun.amt.ru [212.111.64.19]) by amtsun.amt.ru (8.8.8/8.7.3.1) with SMTP id TAA14815 for <nanog(a)merit.edu>; Thu, 27 May 1999 19:38:21 +0400 (MSD) Delivered-To: nanog-outgoing(a)merit.edu Message-Id: <2.2.32.19990527153613.00ad2b70@amtsun> X-Sender: zinin@amtsun X-Mailer: Windows Eudora Pro Version 2.2 (32) Mime-Version: 1.0 Date: Thu, 27 May 1999 19:36:13 +0400 To: nanog(a)merit.edu From: Alex Zinin <zinin(a)amt.ru> Subject: OSPF multi-level hierarchy: Necessary at all? Sender: owner-nanog(a)merit.edu Precedence: bulk Errors-To: owner-nanog-outgoing(a)merit.edu X-Loop: nanog Content-Type: text/plain; charset="us-ascii" Content-Length: 764 Hello, We're currently discussing necessity of multi-level hierarchy in OSPF on the WG mail list. The idea is to implement SPF-based interarea routing with more than two levels of topology abstraction and route aggregation (we have two levels in OSPF at the moment level1 being intra-area routing and level2 being the inter-area one). I have some thoughts on how this could be done, but the main question is whether there is a demand for it or not. Everyone is really welcome to share opinions. Thanks in advance, - - ------------------------------------------------------------------ Alex D. Zinin, Consultant CCSI #98966 CCIE #4015 AMT Group / ISL Cisco Systems Gold Certified Partner http://www.amt.ru irc: //EFNET/#cisco, //irc.msn.com/#NetCisco [Ustas] - ------- End of Forwarded Message ------- End of Forwarded Message

1 0

Re: OSPF multi-level hierarchy: Necessary at all?
by Vadim Antonov 27 May '99

27 May '99

Alex Zinin <zinin(a)amt.ru> wrote: >We're currently discussing necessity of multi-level hierarchy >in OSPF on the WG mail list. ... >but the main question is whether there is a demand >for it or not. I guess the answer is pretty much not. The amount of interior routing information in a properly designed backbone is quite small even if there is no two-layer abstraction. (Now, the _exterior_ information is aplenty, but OSPF is useless for it anyway). If someone proposed simplifying and cleaning up OSPF i'd be quite for it. Building a protocol allowing to get rid of iBGP hack would also help :) --vadim

2 1