May 1997 - Test - lists.nanog.org

Re: UUNET Press Release on Peering
by smd＠clock.org 13 May '97

13 May '97

Peter Galbavy <peter(a)wonderland.org> writes: > Scenario (a) - Demon Internet install a trans-US DS3 backbone, all centred > around New York, since this is where all our traffic flows throw. We have > no US customers, so the network is wasted, and the DS3's get used as > unidirectional pipes to the various NAPs. > Anyone wanna share this network ? In fact branching out in to the U.S. is probably an excellent idea for non-U.S. networks, particularly those with in Asia and Europe with adequate money to invest, ready access to trans-oceanic bandwidth, and subsidiaries or potential partners in the U.S. to handle operational and establishment issues. This is, after all, apparently what BT is doing, and they don't appear to be the only ones. Given the intuitive belief that market share is important to networks who want to be non-local in scope, and the geography of Europe compared to that of the U.S., particularly with the price of unlit or semilit fibre runs about to drop through the floor in some places, this makes perfect sense. Of course, as you half-indicate, this is very expensive for smaller organizations, however it often appears that consortia of smaller regional organizations remain fairly stable in Europe. A hierarchical approach given Europe's geography has some obvious distinct advantages over a centralized "global domination" approach, provided it can remain nimble and flexible enough to satisfy its components, offer good service and continue growing. > Guess which one I like ? But I will settle for (c). (Oh (d) for those who > didn't guess). If Europeans and Asians and Austrialians made strong efforts to keep the amount of traffic being sucked towards them from the U.S. with things like web caching hierarchies (haha, enjoy an international aspect to copyright legislation battles) and mirroring, particularly when this can be done without the total cooperation of end users, then they would be under less pressure to acquire more trans-Oceanic capacity from the U.S. Because culturally this would be much more difficult to accomplish in the U.S., Americans may well start clamouring for more bandwidth to satisfy them. Happy non-Americans, unhappy Americans, cash for increased bandwidth flows from American pockets. Alternatively, this would be great leverage for the kind of scheme you described as (a). Sean.

2 1

Re: UUNET Press Release on Peering
by Sean Donelan 13 May '97

13 May '97

>Strikes me that the terms "diversely routed" and "geographically diverse" >can be interpretted wildly. You catch on quick. Who gets to do the audit? Other than the 'public' interface between the providers, I don't think it is wise, or appropriate to get into the internal design or business plans of competitors. You really have to treat the rest of it as a black box. It hard enough to predict what technology I'll need to deploy in my network without having by hands tied by an agreement with a competitor. Thou shall use only dedicated, fully-meshed, terrasterial clear-channel lines sounds great until I discover the best way I can reach timbuktoo is over a ATM satellite link. DRA doesn't have large Web farms, but people build public libaries in the darndest places. I'm always amused by providers that have peering requirements their own networks, or existing 'peers' don't meet. >For most networks, I don't see a real problem dropping a couple extra >pairs of DS3s (assuming they were only at two NAPs to begin with). There is a partition problem. Assume there are six points, and each provider chooses four. Provider 1 could choose 1, 2, 3, 4. Provider 2 could choose 3, 4, 5, 6. Although each provider is at four points, they only share two in common (3, 4). Which provider alters their business plans to connect to eight places just so they have four common exchange points? >If UUNet requires a DS3 interconnect at LINX, MaeWest, MaeEast and say >(for giggles) Australia -- AND a diversely routed DS3 backbone in the >U.S. that takes on an entirely different meaning than the above. Don't giggle, DRA has an office in Melbourne. We don't have enough traffic to justify multiple DS3/E3's to Melbourne, but looking at UUNET's international map, neither do they. Every provider has their own strengths and weaknesses, they are never exactly equal. Getting into the internal network design of your competitor is just asking for trouble. PSI and Sprint engineers used to battle over who had the better backbone design. Neither ever convinced the other they were right. I don't think its a good idea to require every backbone follow the exact same design. We may have different 'major' routes than other providers, and choose to deploy our 'big' links differently. That's one of the strengths of the Internet, not a weakness. My design methodolgy may be good, or it may be bad; but the design diversity between providers hopefully means a "group think" mistake won't kill everyone. -- Sean Donelan, Data Research Associates, Inc, St. Louis, MO Affiliation given for identification not representation

5 5

ISP tutorial in Tampa
by Susan R. Harris 13 May '97

13 May '97

Greetings - for those of you attending the Tampa NANOG meeting (http://www.nanog.org/mtg-9706/update.html) just wanted to let you know that on Wednesday evening, June 4, Michael Dillon, a well-known systems integrator and ISP consultant, will present an introductory tutorial titled "Review of Internet Resources for ISPs." The session will cover IP address allocation issues, with information targetted especially for providers who are applying for their first IP address allocation from ARIN/the InterNIC. He will also give a review of websites and FTP sites that provide information on: - Internet routing and addressing - Federal and industry network initiatives - Internet standards and other issues that require community cooperation. The tutorial will run from 7:30 - 9:00 p.m. To attend, you must also be registered for NANOG. If you plan on attending, please send a message to Pam Ciesla (pam(a)merit.edu) with the word 'tutorial' in the header. Several tool demos will be presented concurrently with Dillon's tutorial. Hope you can join us for these sessions! ------------------------------------------------------------------------ Susan R. Harris, Ph.D. Merit Network, Inc. srh(a)merit.edu Phone: (313) 936-2100 Fax: (313) 647-3185 ------------------------------------------------------------------------

1 0

Re: UUNET Press Release on Peering
by Sean Donelan 13 May '97

13 May '97

>This is a serious misunderstanding of the economics of Internet. > >The "equity in traffic" is not important at all. > >What is important is equal share in carrying packet flows, which >can be approximately measured in miles*bps. *That* costs money. > >Therefore, a small ISP dumping most of its traffic to others at >one exchange point will cause long-haul backbones to carry its >traffic most of the distance (no matter which direction the traffic >goes), thus effectively subsidizing that small ISP. Like I said, people build public libraries in the darndest places. How would you catagorize an provider that serves customers in 38 states and two canadian provinces. It works out to be about 11% of the public library patron population in the USA. Most of the customers are located in fairly remote areas as far as telecommunications infrastructure. Which means the price of a T1 local loop can cost as much as some providers, who concentrate on more high-volume business customers, pay for a DS3 between larger population centers. Cream skimming comes in all shapes and forms. Figuring out who is 'free-loading' isn't alway clear-cut. Especially when the provider that has the web farm in Virginia, and the provider with the users is hauling the traffic to 38 states and 2 canadian provinces on their backbone. -- Sean Donelan, Data Research Associates, Inc, St. Louis, MO Affiliation given for identification not representation

1 0

RE: UUNET Press Release on Peering
by James Saker 13 May '97

13 May '97

Wayne Bouchard (web(a)primenet.com) writes: >Ultimately, I think farm.com would have to >give in to keep customers from moving services to a "better connected" >provider. Or UUNET customers moving to other "better connected" NSPs, like PSI, MCI, BBN, etc. This too resolves the problem on both ends. Any word on others following in UUNET's footsteps? James R. Saker Jr. Intellitek Inc. President, Network Media Division voice: 402.333.6233 jsaker(a)intellitek.com fax: 402.333.6432 http://intellitek.com

2 1

RE: Coincidence...
by Jim Browning 13 May '97

13 May '97

>In any event, let me just say that I disagree with you >that "stringing is easy" -- replacing all the crap in the >ground in the U.S. now with fibre that has decent >chromatic dispersion properties and as little dependence >on electro-optical conversion as possible is going to be a >long, expensive and labour-intensive process. > >Hence the window of opportunity for Hans Werner von Braun >and his amazing rocketships of death, and similar >techologies I'd rather not have to depend on. :-) I keep thinking that, as costly as it is to upgrade technology in the ground or under a few thousand feet of seawater, those difficulties pale in comparison to what will be entailed in upgrading satellites. Given the rapidity with which networking technology is advancing, one must admire their willingness to take risks...

1 0

Re: UUNET Press Release on Peering
by William Allen Simpson 13 May '97

13 May '97

In view of the difference, I would suggest that the web farms have a case that UUnet should pay _them_ for the priveledge of accessing their content! Anytime UUnet asks for a fee for peering, just tell them that you really consider them unequal, and that they should pay YOU! > From: Mark Turner <markt(a)nsipo.arc.nasa.gov> > For NASA this is a current snapshot of NAP traffic > exchange: > > mae-w: > > 5 minute input rate 2888000 bits/sec, 632 packets/sec > 5 minute output rate 2949000 bits/sec, 718 packets/sec > > mae-e: > > 5 minute input rate 5458000 bits/sec, 2437 packets/sec > 5 minute output rate 8041000 bits/sec, 2157 packets/sec > > The flows are constrained by an over-worked backbone, > but as a general rule out > in. As I said, web-farms > and a small backbone will give you that :-) > WSimpson(a)UMich.edu Key fingerprint = 17 40 5E 67 15 6F 31 26 DD 0D B9 9B 6A 15 2C 32 BSimpson(a)MorningStar.com Key fingerprint = 2E 07 23 03 C5 62 70 D3 59 B1 4F 5E 1D C2 C1 A2

2 1

Re: Coincidence...
by smd＠clock.org 13 May '97

13 May '97

bmanning(a)ISI.EDU writes: > If the PSTN is the only game in town The PSTN is dead; the only thing is that the body is still giving off lots of heat so fibre owners haven't begun giving up on it yet. Some of them even appear to be betting that in three days it'll get up and start walking around again. Real money is in computer mediated and computer-to- computer communications, and it's that money which is sustaining the last vestiges of metabolism in the PSTN. > then its ATM or nothing.. and ATM won't cut it at those > rates. Well it's not surprising that telcos still hope to save the PSTN by selling only ATM like in silly cases where a clear-channel colocation-facility-to-colocation-facility SONET run is ignored as a sales option while an ATM PVC of the same general bandwidth takes a thousand-mile detour to and from the nearest ATM switch... Moreover, I'm told that currently on offer from one telco is a OC3 SONET pipe for only ten times the price of a DS3. However, if you want a PVC following the exact same path, the circuit is only a little more than double DS3. (Ironically you can currently do 4:3 SONET/SDH inverse muxing giving you an OC3/STM-1 from 4xDS3 (plan your DS3s along divers paths, kids!). Of course, you can just use various router vendors' strategies for load-balancing among DS3s and use all 4 DS3s as they are, for that matter...) Finally, to tie in a reply to your reply to my previous message, finding ways to light up fibre is not so difficult; at worst it all comes down to massive amounts of MUXing. While I happen to prefer the approach of a combined SONET MUX/IP router to stick at each end of a pair of soliton-containing fibre strands, Vadim Antonov's approach is also tractable and attractive, and has some interesting benefits that could be seen as side effects wbut which he has turned into a real additional-value feature within his proposed product. In any event, let me just say that I disagree with you that "stringing is easy" -- replacing all the crap in the ground in the U.S. now with fibre that has decent chromatic dispersion properties and as little dependence on electro-optical conversion as possible is going to be a long, expensive and labour-intensive process. Hence the window of opportunity for Hans Werner von Braun and his amazing rocketships of death, and similar techologies I'd rather not have to depend on. :-) Sean.

1 0

Re: Coincidence...
by smd＠clock.org 13 May '97

13 May '97

bmanning(a)ISI.EDU writes: > And its not clear there is enough > capacity in the PSTN to carry all the bits about. The hard part is stringing fibre around, followed by figuring out clever ways of lighting it up. North America is Geographically Challenged in that respect compared to the greater number of shorter rights-of-way among large customer bases in other parts of the world. For now there is some coasting going on in the U.S. and Canada particular since the traffic crunch is mostly north-to-south. However ultimately trunking east-to-west will fill and that is a much harder bottleneck to widen, and provides interesting opportunties to people deploying things other than fibre in the U.S. Of course, such deployments strike me as unstable economically principally for scaling and failure-avoidance-and-recovery reaosns compared to modern fibre technology. Of NANOG relevance, there are ample war stories of weather-sunspot-and-bat-releated outages on high-bandwidth alternatives to terrestrial paths. *I* wouldn't use a non-fibre (and ideally non-SONET) path where one were avialable if losing connectivity (or alot of capacity) when it stops working were an issue. Sean.

2 1

Re: UUNET Press Release on Peering
by smd＠clock.org 13 May '97

13 May '97

Peter Galbavy <peter(a)wonderland.org> writes: > I am using UUNETs policy as an example of why some of > these do not make sense in a global scene - yes I also > know this is NANOG) You can hardly lessen the signal to noise ratio here. I think UUNET's policy is fundamentally sound technically, and I applaud it. This should surprise noone, because it's what I was trying to accomplish rather publically before my disappearance. However, UUNET seems to have gagged its technical staff who undoubtedly could do an infinitely better and clearer job of explaining their policy in this kind of forum than their recent press releases have done. On the other hand, this may be a pragmatic business decision, since frankly, this forum is completely irrelevant to their business strategy (and to almost everything else, these days), and any real concerns probably tend to deal more with misperceptions potential customers might have about UUNET's connectivity, capacity, and overall utility. Froma business perspective, those sorts of concerns probably are best addressed by P.R. folks. I note that their press releases have also made much mention of UUNET's forthcoming $300M investment in infrastructure. This is hardly coincidental. It's also about time that someone put some real money and an apparent strategy into the Global Internet. Sean.

1 0