- NANOG - lists.nanog.org

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

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

13 May '97

"William Allen Simpson" <wsimpson(a)greendragon.com> writes: > 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! This would be a really good idea because it would give a strong push into investing in web caching technologies at the ISP level. Once that's been done, and access-level devices (customer-aggregating NATs and the like) can be gotten reliably to intercept queries directed towards anything other than a web caching hierarchy, in an intelligent engineering effort to keep traffic as local as possible when it's possible, the whole spurious argument that this is about web farms will go away. Moreover, a more general piecemeal settlements scheme would be much easier to put in place in an environment where the web farmers or small providers argue back that they ought to be paid by larger providers with huge numbers of end users connected through them. Finally, if you were UUNET and looking at things from a UUNET perspective, what's the difference between an entity exchanging traffic with them in one or two places across an FDDI switch and an entity exchanging traffic with them in one or two places across a Frame Relay switch? And wouldn't you start wondering when small providers will start paying their own customers for the privilege of talking with them? Sean.

1 0

RE: UUNET Press Release on Peering
by Peter Galbavy 13 May '97

13 May '97

On Monday, May 12, 1997 6:53 PM, Deepak Jain [SMTP:deepak@jain.com] wrote: > Strikes me that the terms "diversely routed" and "geographically diverse" > can be interpretted wildly. > > 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 if you do not have any "domestic" US customers. See below. > 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. ... Cool. I can do the LINX. :) Only problem here for me is that (miles * BPS) is a formula that does not work. My company has a very large customer base in the UK and is a direct competitor of UUNET PIPEX. Big deal you say ? Networks who are competitors always interconnect in the US you say ? This doesn't work in the big picture for us, sitting and the end of a very expensive international T3 line and only competing in one market area. (BTW We have not actually ever requested peering from UUNET outside of the AS1849 seen at the LINX - so this is all hypothetical and 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) Watch my hands on this one, and I will try to go slowly. 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 ? Scenario (b) - Demon Internet "pay" a direct competitor in their own market for transit. I hope I do not even have to spell out the wonderful sales calls our customers and prospects would get ? (BTW We *pay* MCI for >transit< services and have come to an arrangement with PSI - for those who may try traceroutes - but until the Concert merger is final, MCI is *not* seen as a competitor - and PSI we can deal with, since we are all nice people). Scenario (c) - Demon Internet approach a number of large NSPs in the US asking about customer-network peering only (not transit) at some mutually agreed single point - sharing the actual costs of delivery. In the US of course. Scenario (d) - Demon Internet approach same people as in (c) but for two points, one in our territory - the UK. This means the US NSP has an international backbone to worry about. Guess which one I like ? But I will settle for (c). (Oh (d) for those who didn't guess). Anyone out there who want to *peer* with us in Scenarios (c) or (d) who does not already do so, please let me know in private e-mail. Anyhow, the point I am trying to make is that it is very much pointless for companies like us to build a backbone when I do not have customers to use it, and also that paying a company in the US for services, when we have an "aggressive" competitive relationship in our own market is a non-starter. Regards, -- Peter Galbavy Demon Internet Ltd http://www.demon.net/

1 0

Re: NSI and competition to RIPE and APNIC
by Carl Oppedahl 13 May '97

13 May '97

At 10:22 AM 05/12/97 CDT, Stan Barber wrote: >Carl writes: >> And the outcome of all this affects every Nanog member. It affects every >> network operator, every ISP, every sysop. It affects the 1.2 million >> owners of the 1.2 million domain names in COM and NET and ORG. Every >> network operator, every SIP, every sysop, every domain name owner, should >> pay attention to all of this and speak up. > >Carl, I don't think there are 1.2 million domain name registrants as many >domains (like CyberPromo, for example) has registered multiple domains names. Nowadays, taking annual fees into account, the number of parties owning thousands of domain names is dropping quickly. Okay, so let's say it is a ten-to-one ratio (which it's not, I believe it is more like 1.5 to one) then the number of domain name owners would still be 120,000. And those 120,000 owners should still be speaking up, in the absence of which things may happen that they won't like later. >That being said, there is already an alternative in the county code domains. >There, the policies are different (though not necessarily better) than NSI. That is no comfort at all, none whatsoever, to someone who has invested years of time and sweat and energy and money in a COM domain name. If the answer is, "Hey, you don't like NSI's domain name policy? So give up your domain name and start over again in a two-letter domain!" then it is no answer at all. >That's not to say that the exising gTLD infrastructure does not need to >evolve. It must. I agree with you that those who are domain name registrants >in the current gTLDs have a stake in this evolution and should express a >thoughtful and constructive opinion on how this evolution should take place. >Unfortunately, some who have expressed opinions have chosen not to be >either thoughtful or constructive. I think some operators have been reluctant >to participate because of this. As I am sure you are also aware, the present structure of the .US domain is unworkable for any business that is located in more than one state, or for any business that thinks it might ever move from one state to another. There ought to be a .com.us, for example, and at present there isn't one.

3 2

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

13 May '97

>Do you feel that UUNET's own network doesn't meet the criteria set out >in the press release? How so? > >Which of UUNET's peers who are able to continue to peer with them, would >you say do not follow the criteria set out in the press release? Actually I was thinking of a few other providers, but it is a common occurance among all providers with requirements for peering. MCI's requirement for a fully-meshed backbone. AGIS's requirement to connect to five exchange points, and already have peering agreements with many other providers. I haven't seen UUNET's actual peering requirements, just the press release, so maybe this is just a case of poor PR writing. But the press release does give some details about the backbone. It says DS3's or higher are used on "major" routes. Which begs the question, what does the backbone use on "minor" routes? Few providers have homogenous backbones. They usually have a wide range of circuit types, speeds, media. The definition of backbone, major, route, diverse, dedicated, etc are ratholes you go down when you venture past the 'public' interface between providers. I'm not trying to pick on a particular provider, just pointing out things you need to consider if you plan to write a requirements document. I don't like putting ratholes into requirements if they aren't really requirements. Last summer, you could substitute AGIS for UUNET, and had the same debate. In the end its not a techinical decision, its a business decision. I would really like to avoid the fiasco that happened with Western Union and the whole notion of "International Record Carriers" and "Record Carriers" that happened before the FAX machine made the teletype obsolete. -- Sean Donelan, Data Research Associates, Inc, St. Louis, MO Affiliation given for identification not representation

1 0

Re: UUNET Press Release on Peering
by Vadim Antonov 13 May '97

13 May '97

On Mon, 12 May 1997, Christopher Morrell wrote: >> Which of UUNET's peers who are able to continue to peer with them, would >> you say do not follow the criteria set out in the press release? >I know of at least one. I'm sure there are more. It seems, >that equity in traffic exchange may be an overriding >principle, although running many web-farms and small >backbone can certainly give you that. 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. --vadim

5 5

Re: REMINDER
by Dimitri Krinos 12 May '97

12 May '97

In Pam's absence today, let me just mention that there are still a few rooms left at the conference rate for NANOG 10. The cut-off for the conference rate ($129) is 6:00 PM Est on May 15th. After that, the Hyatt's standard room rate for single occupancy is $205/night. If you do make reservations after the cut-off date, please still mention NANOG so that you get your conference documentation, etc. Dimitri Krinos IBM Global Services At 08:57 AM 5/9/97 -0400, Brad Miller wrote: >There are no rooms left at the discount rate. The current rate is $195.00. > Wow. > > >At 08:36 AM 5/9/97 -0400, Pam Ciesla wrote: >> >>The Hyatt Regency room rate cut-off date is May 15. The phone number >>is 1-800-233-1234 or (813) 874-1234, be sure to mention NANOG. The >>attendee list has been updated. If you are not on the list and should >>be please contact me. If you are planning on calling with payment info >>please note that I will be out of the office Monday, 5/12 & Tuesday, 5/13. >>Registration fee of $100.00 is due by May 28. Cost after the 28th is $120.00. >> >>xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx >>Pam Ciesla 313-936-0172 - phone >>Merit Network, Inc. 313-647-3185 - fax >>University of Michigan >>4251 Plymouth Road >>Bldg. I, Suite C >>Ann Arbor, MI 48105-2785 >>xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx >> > > > > >____________________________________________________________________________ >___________ >Brad Miller 703-448-4470 ext. 424 >Director of Operations bmill(a)cais.com >CAIS Internet > >

2 1