- NANOG - lists.nanog.org

Re: Provider credibility - does it matter? was Re: Inter-provider relations
by Bill Cessna 28 Oct '96

28 Oct '96

>>From: jakob(a)jubii.dk (Jakob Faarvang) >>To: "Alex.Bligh" <amb(a)xara.net>, roll(a)stupi.se, nanog(a)merit.net >>Subject: Re: Provider credibility - does it matter? was Re: Inter-provider >relations >>Date: Sun, 27 Oct 1996 14:42:43 GMT >>Organization: cybernet.dk >>References: <199610271111.LAA11077(a)diamond.xara.net> >>Sender: owner-nanog(a)merit.edu >> >>On Sun, 27 Oct 1996 11:11:10 +0000, you wrote: >> >>>> >>>> I don't think this is correct. Ebone peers with other pan-european >>>> networks of comparable size, IBM, BT and EUnet comes to my mind. >>>> >>>In fact, last time I heard, EBONE and EUnet's peering policy >>>bore an uncanny resemblence to those of Sprint & MCI (albeit >>>bandwidth scaled). >> >>From Ebone's own pages, BT is listed as a customer to Ebone, and it is >>said the IBM and EUnet have transit agreements with Ebone. I can't >>find any direct text stating Ebone's peering policies, but on the >>Amsterdam IX page (http://www.ams-ix.net/) the following is said: >> >>Policy: Does generally not peer except with other larger ISPs after >> negotions. Become a member and we'll peer with you! >> >>- Jakob Faarvang >> >> IBM Global Network peers with EBONE. We do not get or buy transit from EBONE. The EBONE home pages (last time I checked) did not provide enough information to crisply tell who were members and who were peering partners. To peer, you must connect with EBONE at three places at E-1 speed as a minumum. Web pages for EBONE are at (http://www.ebone.net/) and say this if you read carefully. (I think they do...) Members get transit. That is what being a member of EBONE is about and how the EBONE backbone provides value to its members (members can be thought of in terms of a customer). It is a model that works well in Europe and provides for all ISP's to join. Fees that support the EBONE backbone are based on the members access to the EBONE backbone line speed. Bill Cessna Bill Cessna, IBM Global Network, Network Architecture Internet: wcessna(a)vnet.ibm.com Internal: IPNET(WCESSNA) Voice: +1-813-878-3947 Fax: +1-813-878-4250

1 0

US transit
by Per Gregers Bilse 28 Oct '96

28 Oct '96

I'd like to set one thing straight: It is completely bogus if people run around believing there are huge amounts of non-US traffic transiting the US, clogging up exchanges and whatnot. The volume of this kind of traffic is completely in the noise compared to traffic which is sourced and/or destined in/for the US; any argument suggesting the opposite is misguided or wilfully misleading. -- ------ ___ --- Per G. Bilse, Mgr Network Operations ----- / / / __ ___ _/_ ---- EUnet Communications Services B.V. ---- /--- / / / / /__/ / ----- Singel 540, 1017 AZ Amsterdam, NL --- /___ /__/ / / /__ / ------ tel: +31 20 5305333, fax: +31 20 6224657 --- ------- 24hr emergency number: +31 20 421 0865 --- Connecting Europe since AS286 --- http://www.EU.net e-mail: bilse(a)EU.net

1 0

Re: Provider credibility - does it matter? was Re: Inter-provider relations
by scope＠cais.cais.com 28 Oct '96

28 Oct '96

At 06:28 PM 10/28/96, Jim Dixon wrote: >On Mon, 28 Oct 1996, Jian Li wrote: > >> > The point is that GSL's peering policies are radically different from >> > those of SprintLink, whereas Cable & Wireless seems to have the same >> > policies worldwide. Similarly, I know that GSL's prices for transit >> > used to be much higher than SprintLink's and I suppose still are. >> >> As what I know, GSL's peering policies have two parts, international and >> US domestic. The US domestic part is exactly the copy of SprintLink. >> International part is based on the idea of traffic localization. > >Fine. This still leaves Sprint with two very different policies, one >for the US and one for the rest of the world, whereas CWIX apparently >has one policy for the whole world -- or at least for North America and >Europe. I believe that over the next few years this is how things are >going to go: peering policies and prices for transit will converge, at >least in these two large markets. Sprint has appearently made the decision to treat Sprintlink and GSL as two entirely seperate networks. This makes sense based on the way that they structured Global One. The GSL product is appearently intended to be an international Internet product. In this sense they really look more like a no U.S. provider with the rate structure and cost basis to match. If the idea is to drag the traffic to the U.S. (whether because it is cheaper to do so, U.S. traffic is a bigger draw, or the U.S. telephony infrastructure and price structure causes the U.S. to be attractive for Asia - Europe transit) then GSL has to price accordingly. Sprintlink, being primarily focused on the U.S. domestic market has a much easier time with underlying costs and customer concentration. Their terrestrial fiber (as Peter L. pointed out) makes it so. > >Americans tend to forget that Europe has a larger population and more >money than the US; these basic facts are not going to let the current >large imbalances persist. The imbalances will fade when and if Europe takes its huge population and bigger bucks (marks, pounds,francs,ecus) and wields them like one economy, not a dozen. What drove the U.S.centric model in the first place wasn't lack of European peering, it was that, way back when, It was cheaper to get to Paris (or name your city in Europe) via Washington than straight on. j >-- >Jim Dixon VBCnet GB Ltd +44 117 929 1316 fax +44 117 927 2015 >http://www.uk.vbc.net VBCnet West +1 408 971 2682 fax +1 408 971 2684 > > Jeff Studds Internet Stuff scope(a)cais.com phone +1 301 699 1840 The internet: It's not just an adventure... it's a job:-)

1 0

Cisco Memory.
by plynch＠jefferson.ind.net 28 Oct '96

28 Oct '96

Nathan, I have to follow up on this comment you made. I work for an ISP in Indianapolis called IHETS. We have 7 cisco 7500 series routers on our backbone. Wre just spent a chunck of money to upgrade the RAM on them to 64Meg. If yoiu are saying that regular SIMMS work in Cicso routers, then I'll shoot myself in the foot. I'd like to know if you have used this method and how well it works out in the field. Does latency increase with *unqualified* Cisco memory, or is the *unqualified* *qualified* Cisco ram stuff just a big ruse? Thanks for your input. > > On Mon, 21 Oct 1996, Alexis Rosen wrote: > > > alex(a)relcom.eu.net writes: > > > and WHY have I to pay 3,500$ for the 32Mb ram if this RAM costs > > > 600$ on the free market???). > > > > Dunno for sure about the 4xxx, but the 7xxx use standard SIMMs, and Cisco > > has even qualified some manufacturers' parts. Oddly enough, they have > > also *disqualified* some parts that you'd think were exactly the same. I > > don't know the details there. The sales types obviously don't like to > > talk about it, but you really *can* get $12k of 7500 memory for $800. And > > they'll even agree that it works, if you talk to the right cisco folks. > > Yep, it uses standard SIMMs, I have upgraded many of them.

13 20

Re: Provider credibility - does it matter? was Re: Inter-provider relations
by azeem＠dial.pipex.com 28 Oct '96

28 Oct '96

>Karl Denninger <karl(a)Mcs.Net> wrote: > >>Any provider that does not recognize the value of bilateral, no-settlement >>peering anywhere that its cost-effective for both parties (ie: if you have >>traffic destined for me, get it on MY network where I'm being paid to >>carry it and let ME figure the rest out!) deserves what they get. > >Zero-settlement peerings open to anyone are demonstrably amount to >subsidies from large peers to small. Absolutely correct: can I direct you to two of our articles on the subject (and a useful graphic): http://www.economist.com/issue/19-10-96/ld4401.html http://www.economist.com/issue/19-10-96/sf0774.html http://www.economist.com/issue/19-10-96/sf1.gif >Zero-settlements work only when peers are of comparable size. >Any attempt to extort pressure to force it upon anyone simply >causes large folks to flee. That's true: in a conversation with Phil Lawlor (CEO of AGIS), he told me that one of his customers ISPs, who was also peering with AGIS at the Mae-E, indulged in "bandwidth stealing" by receiving their Usenet news feed across Mae-E rather than as an AGIS customer across the few T-1s they had to AGIS. Surely a hierarchy which implements the rule you outline above works: a pair of small regional ISPs may want to peer regionally for good reason, because they are the same size there isn't much inequity in this relationship. Two large ISPs may wish to do the same. However, a small ISP who wants its traffic carried by a large ISP (i.e. peer with them) may have little choice but to pay for this traffic to be carried: so becomes a customer. That way you have a hierarchy as follows: Bacbkbone providers who may peer with each other Large regional providers who are customers of the backbone providers (so by DS-3s or whatever from BBNplanet, UUnet) and who may peer regionally. While also present at bix exchanges, they are unlikely to peer with the big six/seven/eight but may well peer with foreign ISPs at those exchanges. Azeem Azeem Azhar vx: 0171-830 7133 The Economist fx: 0171-839 2968 25 St James's Street London SW1A 1HG

14 39

Re: October 1996 Meeting Notes
by sob＠academ.com 28 Oct '96

28 Oct '96

That's because I said the notes "would be appearing" this week. They have not appeared yet. -- Stan | Academ Consulting Services |internet: sob(a)academ.com Olan | For more info on academ, see this |uucp: {mcsun|amdahl}!academ!sob Barber | URL- http://www.academ.com/academ |Opinions expressed are only mine.

1 0

October 1996 Meeting Notes
by sob＠academ.com 28 Oct '96

28 Oct '96

They will start appearing this week at this URL: http://www.academ.com/nanog/oct1996 Enjoy. Comments welcome.

2 1

RE: Provider credibility - does it matter? was Re: Inter-provider relations
by Hank Nussbacher 28 Oct '96

28 Oct '96

On Thu, 24 Oct 1996 18:26:12 -0400 (EDT) Gordon Cook wrote: >Cook: six global providers: Hmmm. Might they be referring to Sprint, AGIS >UUNET, PSI, MCI, and ANS? The so called gang or club of six that a year >ago were the exclusive direct peers at all the major exchanges? If they >are, then, by process of elimnation, BBN is not a global provider -which I >think is news to BBN, nor is AT&T, nor is Advantis, nor is Compuserve, nor >are a lot of other major players with more than a fresh12.5 million behind >them. Gee. I wonder if BBN realizes that AGIS has relegated it to second >rank status? > >Or perhaps they are referring to those global providers who are beginning >to move most of their traffic through private inter connects - and in so >doing are becoming a new internet apex? In this case they are not even one >of the five let alone six. Here the five are MCI, Sprint, BBN, UUNET and >ANS. I have been talking to a lot of sources, none of which is aware of >AGIS having any private Interconnects with these five. When talking of global Internet reach, the largest is IBM/Advantis IGN. They have 830 POPs in over 50 countries and peerings at about 15 major interconnect points. For details see: http://www.ibm.com/globalnetwork/inetbbon.htm Probably the best kept secret in the Internet world. Hank Nussbacher IBM Israel

3 2

RE: Airplane crashing into Atlanta-NAP
by owen＠DeLong.SJ.CA.US 28 Oct '96

28 Oct '96

> On Sun, 27 Oct 1996, Chris A. Icide wrote: > > > Now that I've thrown in my share of late night sarcasm, It would interest > > me greatly to understand exactly why you came to the following > > conclusions: > > > > 1. The "operation" in Atlanta is Mickey Mouse > > Both the Internet and the POTS system are telecommunications networks that > are vital to the modern economy. POTS moreso than the Internet right now > but the Internet is certainly heading in that direction. POTS exchanges > are always in ground floor concrete buildings with no windows or > underground. But in Atlanta they stick it up on the 5th floor of some > office building?!?!?!? > I do not know where you get your data, but you are wrong. Many POTS exchanges in California, and many other parts of the country are not on the ground floor, or in basements. In fact, several of the POTS exchanges in San Francisco are on multiple floors of the same building. > > 2. The floor # of a bulding affects the quality of the Exchange Point > > 3. The type of building affect the quality of the Exchange Point > > I cannot explain these ones but have reached this conclusion from > observing how the phone company builds and locates its exchanges. > Must be specific to your phone company or an incomplete perspective of phone company methods. The primary reason for telephone exchanges of the past being built on ground floors and basements was the weight of the equipment and the difficulty of moving it. In modern times, the weight of the batteries still plays a role, but exchanges are being remoted away from the batteries with increasing regularity. > > 4. A city the size of Atlanta needs more than 1 Exchange Point > > It has a lot more than one POTS exchange. Thus it will need more than > 1 Internet exchange. Why should the packets from every video-call in the > city all travel downtown when frequently the two parties live in the same > neighborhood? > An interesting theory, but the Internet does not route information the same way the telephone system does. For one thing, you are comparing a connection oriented network to a connectionless network. There are subustantial differences. > > Truly I would be very interested in your thoughts on these items, as > > well may a few other folks on this list. > > I'm taking a long term view in which ISP's are just another form of > telephone company. Many ISP's are now getting to the size where they can > consider aquiring strategically located properties, building concrete > block exchange/colo buildings, wiring up entire office towers with > IP fiber and even running their own fibre in some case, especially in new > subdivisions. > ISP's will never be another form of telephone company. ISP's who are viewing their role that way will not last long in the future. Even telephone companies will not maintain that role for very many years to come. You are trying to take a technology of the future and apply it to an obsolete model. > By today's standards Atlanta-NAP may be a really great thing, but too soon > we will discover that we aren't living in "today" any more and the > standards will be different. > Yes, they will, but I don't think they will go back to the Telco of the '50's. > Michael Dillon - ISP & Internet Consulting > Memra Software Inc. - Fax: +1-604-546-3049 > http://www.memra.com - E-mail: michael(a)memra.com > > Owen

3 2

Re: the Union makes us Strong
by Vadim Antonov 28 Oct '96

28 Oct '96

William Allen Simpson" <wsimpson(a)greendragon.com> wrote: >Several have noted in the past that Vadim has a religious predeliction >toward "robber baron" oligopolies. We've had plenty of historic >experience with them here in the US. Hm. You never tried to live in a place without antitrust laws. I bet you'd have quite different perspective then. >The wolves have learned to eat grass (IP) in addition to their old fare >(overpriced direct links, X.25, and OSI). No, they don't. Not while the bellheads are telco's kings of the dumps. I would venture to guess that telcos could have Internet for lunch if they were willing to sacrifice some corporate sacred cows. I learned that hard way -- remember which telco managed to get viable Internet business first? It wasn't that hard to make some competitors existing at the time very unhappy -- all it took is one engineer who can deal with socialism and one manager who was willing to bend the rules. I could also tell why there's no significant competition from telcos in the dial-up field, but i'd rather prefer not to make more personal enemies. >Gentlefolk, if we are to avoid the doom that Vadim predicts, the >large number of rabbits need to band together to fend off the wolves. No. Then "rabbits" will become about as good as the wolves. The role of small business is to move fast and skim the cream. The role of big business is to exploit economies of scale. Internet is no longer an esoteric technology; so i don't see a reason why it cannot be done by telcos. >There was a great deal of talk among small providers in the halls and >restaurants around NANOG last week. Several different models of >cooperation were discussed. Right. How about actually putting fiber in the ground? Unless small ISPs have their own data transmission facilities they have zero chance to win in the long run. >As I raised as an issue with the large providers during their >presentation, the very nature of IP statistical multiplexing improves >throughput when several providers join together and use the same pipe. Done some reading on basics of queuing theory, didn't you? Unfortunately, the theory is dry and tree of life is green -- i hate to point out obvious but stochastic multiplexing does not work for self-similar traffic like it does for Poission. Things are a lot more interesing when you've got traffic with infinite interarrival time variance. No matter how much you aggregate you still get the same stuff. >Yet, if we are to survive, we will need to build a shared cooperative >infrastructure tailored to our own needs. The writing is on the wall. Right on spot. Can I suggest some reading of Lenin? He dwelt quite a lot on cooperation and collectivism. That didn't work in reality, though. >I propose that ISPs join together, build, and fund regional and >metropolitan exchanges, and each of these exchanges make a group >purchase of transit to other exchanges, instead of each ISP purchasing >their own transit. Eventually, the independent exchanges together >will constitute a super provider which meets the draconian peering >requirements promulgated by the big providers to limit competition. You either end up with franchises or a large corporation taking over. Been there, done that. You may want to look at history of Relcom, which was exactly what you descibe there. Didn't last long that way, though. >I believe that we need an umbrella organization to make this happen. >There is apprehension that the CIX might involve conflict of interest >among its membership. The new ISPC might be a good choice. CIX would be better off not providing any service in competition to its members. If an entity is forwarding packets it _is_ in competition with its "customers" or members, unless it is in it's chapter to disallow membership of anybody who pays fees. When you do that it's no different from a cooperative, an economical form known to be unstable and prone to "spontaneous" decay into a conventional corporation. --vadim

1 0