- Test - lists.nanog.org

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

More rats?
by Wolfgang Henke 28 Oct '96

28 Oct '96

traceroute www.cisco.com: traceroute to cio-sys.cisco.com (192.31.7.130), 30 hops max, 40 byte packets 1 192.100.81.254 (192.100.81.254) 2 ms 2 ms 2 ms 2 sjx-ca-gw1.netcom.net (163.179.1.29) 3 ms 7 ms 3 ms 3 t3-1.scl-ca-gw3.netcom.net (163.179.220.194) 244 ms 98 ms 96 ms 4 * mae-w.bbnplanet.net (198.32.136.14) 64 ms 75 ms 5 paloalto-br2.bbnplanet.net (4.0.1.13) 85 ms 52 ms 196 ms 6 paloalto-cr1.bbnplanet.net (131.119.0.201) 51 ms 52 ms 59 ms 7 su-pr2.bbnplanet.net (131.119.0.218) 57 ms 56 ms 56 ms 8 * su-pr2.bbnplanet.net (131.119.0.218) 41 ms !H * 9 * * * 10 * * * 11 * * * 12 * * * 13 * * * 14 * * * 15 * su-pr2.bbnplanet.net (131.119.0.218) 76 ms !H * 16 * * * 17 * * * 18 * * * 19 * * * 20 * * * 21 * * * 22 * * * 23 su-pr2.bbnplanet.net (131.119.0.218) 102 ms !H * * 24 su-pr2.bbnplanet.net (131.119.0.218) 103 ms !H * * 25 * * * ping -s www.cisco.com: PING cio-sys.cisco.com: 56 data bytes ICMP Host Unreachable from gateway su-pr2.bbnplanet.net (131.119.0.218) for icmp from netcom17 (192.100.81.130) to cio-sys.cisco.com (192.31.7.130) ICMP Host Unreachable from gateway su-pr2.bbnplanet.net (131.119.0.218) for icmp from netcom17 (192.100.81.130) to cio-sys.cisco.com (192.31.7.130) ----n-addr.arpa PING Statistics---- 36 packets transmitted, 0 packets received, 100% packet loss Looks like a free day of service for Cisco? Wolfgang

4 3

Re: Atlanta-NAP's choice of switch
by Paul J. Zawada 27 Oct '96

27 Oct '96

At 08:12 PM 10/26/96 -0700, Jonathan Heiliger wrote: >On Sat, 26 Oct 1996, Darin Wayrynen wrote: >What else would you suggest? Gigabit Ethernet hasn't been standardized >yet, Cisco doesn't make a HIPPI interface, and some people prefer to not >use ATM. FDDI has proven to very reliable, etc. Having ISPs continue to >grow egress bandwidth has shown to be a bigger problem than the switch >fabric at the larger exchange points. HIPPI would probably make a poor technology to build a NAP out of. The set-up time that the HIPPI-SC protocol takes to set up a connection introduces too much latency for small packets. While HIPPI works great for large bulk data transfers (using tuned applications, including tuned ftp implementations), I think the performance would be pretty poor when it came to exchanging tinygrams between NSPs. HIPPI might be better than full duplex FDDI, but something like OC12 ATM may be better, even after considering the packet-shredding factor... --zawada Paul J. Zawada, RCDD | Senior Network Engineer zawada(a)ncsa.uiuc.edu | National Center for Supercomputing Applications +1 217 244 4728 | http://www.ncsa.uiuc.edu/People/zawada

1 0

Re: Provider credibility - does it matter? was Re: Inter-provider relations
by Vadim Antonov 27 Oct '96

27 Oct '96

Karl Denninger <karl(a)Mcs.Net> wrote: > >> Zero-settlement peerings open to anyone are demonstrably amount to >> subsidies from large peers to small. >No they're not. The load which the small provider presents to you (in the >form of traffic to your CUSTOMERS) is miniscule by comparison. Small providers outnumber large providers by quite a lot. In fact, most traffic is generated by customers of small ISPs. >Finally, and FAR more importantly, the REASON you're having the traffic >dumped to you is that *YOUR CUSTOMER IS PAYING YOU TO GET IT TO THEM*. The customers are paying for connectivity to other customers, not for connectivity to ISPs. They don't care less if their peer is connected to a small or a large ISP. >If you refuse to perform that job, then your customer should find someone >who will actually live up to the letter and spirit of what your customer >is purchasing from you. I do not think that leftish political activists and radical idealists are a large segment of market. The rest would simply go to a provider who sells better and cheaper connectivity, without worrying too much about figures on the provider's balance sheet. > That already was beaten to death. However, i repeat the argument: > > Big Provider > Customer A ---[POP] ------------- 1000 miles -----------[POP] > | > IXP > | > Customer B ------[POP]-1 mile-[POP] > Small Provider > > When customers A and B talk Big Provider pays to get them through > 1000 miles. Small Provider pays for 1 mile. >So what? Customer A paid you to get the traffic to him. Now, if Big Provider competes for customer B it'll have to sell service for cost of 500 miles, whereas Small Provider can sell is for the cost of 1 mile. The situation is the same at Customer A's location (where Small ISP #2 operates). Now, the traffic between Small ISPs #1 and #2 is close to non-existent; so they simply dump 99.9% of traffic to the Big ISP at exchange points and do transit over a cheap low-bandwidth line they buy from the same or other Big ISP. Essentially, they get benefits of global infrastructure without contributing anything to it. Of course, large ISPs then have to pass the costs to customers, placing them at a competitive disadvantage. >It is in your best interest to do it. You got paid to do this. If you >can't, Customer A will find Big Provider #2 (or Small Provider #2) who >will. That is not happening now. Sprint tested waters back then with CIX, and the expected negative customer response was zero, nada, nil, zilch. Which only confirms that customers do not care about particular ISP's, they only care about bits getting delivered. >The first time you tell a CUSTOMER as "Big provider" that "the reason you >can't reach Customer B, who you think is important, is because they aren't >connected to us and their provider won't *pay us to transport YOUR DATA* you >are going to find out, quickly, what the Customer's response to that is. Huh? The small ISP (Mukhosransknet, for example) who can't reach Sprint or UUNET or BBN is bust by definition. The large ISP who can't reach the same small ISP doesn't suffer at all -- chances are that nobody will ever notice. So the small ISP is forced to buy connectivity to the Big ISP, and the problem (from the point of view of the Big ISP's customers) is fixed. That's particularly true because most small ISPs do not provide any popular content, so customers of big ISPs have little reason to worry about unreacheability of small ISPs . >You might find out from their corporate counsel; if not, you'll definitely >find out from their purchasing department (or person) -- when they cancel >your service and move somewhere else. How idealistic. Customers terminate service all the time for a million of various reasons. Somehow more customers are subscribing, so those departments just don't care. >Second, if Small Provider who *does* have capacity to that exchange point >finds out what you're up to, expect to have that widely used in press >materials and marketing efforts. Just to have it pointed out that this is a free country and that large ISPs have right to make their own business decisions? You cannot make anti-trust case because there _is_ a competition among a number of large ISPs. It's not like one large ISP opressing all others. It's more like reach "oppressing" the poor by refusing to give away their assets to any tramp from a street. Sure, such a press campaighn is possible, and will attract some attention from leftist fringe. The level-headed people would take it to reflect on the respectability and the agenda of the campaighner. > Note that i didn't even talk about less measurabe, but way too > more important things like hosting of information suppliers. > Say, Big Provider connects 1000 web sites; Small Provider hosts > 1 site -- benefit from peering in terms of Web site diversity to > the Big Provider's customers is 0.1%. To Small Provider's > customers the benefit of peering is 99.9%. >Not if you're a customer of Big Provider and want to get there. Your >provider either PROVIDES or you find someone who will. Find me a person who wants to look at Mukhosransk's city council web page. >You seem to forget the middle letter of ISP is *SERVICE*. You want to >talk to someone with a valid IP address on the Internet, your PROVIDER is >responsible for seeing that you can get to an exchange point where can be >found the network that serves them. Ah, what a bunch of rhetoric. >BOTH providers have an obligation here, and its not to each other. Its to >their CUSTOMERS. Obligation to whom? To the Supreme Deity of All Networks? Or the ghost of Comrade Lenin? That obligation is certainly not in service contracts. >If the receiving network then refuses to accept the traffic destined for >a customer WHO IS PAYING THEM TO TRANSPORT IT, the customer on that end >has a very legitimate beef with their provider and IMHO has every right to >walk away and possibly even sue, contract or no (the provider just breached >their material obligations). You may want to take a look at a service contract someday. There's not even a remote hint of implicit or explicit obligation to provide universal connectivity. If _you_ sell service with obligation to provide universal connectivity, i (and my lawyer) want to sign up. He (the lawyer) just loves to shake damages from likes of companies which promise Harriers for $100000. Can you statements in this public forum to be construed as an official position of MCSNET? >Anyone short-sighted enough who fails to understand that bilateral, >no-settlement PEERING (*NOT* transit) is in everyone's best interest >deserves what the market does to them. Sorry, buddy, i'm not interested in everyone's best interest. I just don't care about everyone. I'm admittedly interested in the bottom line in my account statement. I wouldn't expect large ISPs to do differently. After all, they exist to generate profit. If they would start doing charity and brotherhood of all people stuff instead of business their shareholders will be rightfully upset. >>I presume that (1) the people you peer with are clueful and don't do stupid >things on a regular basis, (2) they don't try to point default at you, etc. >That's a given in these discussions. I wouldn't trust people who can't comprehend realities of everyday life to have a clue about routing. --vadim

3 2

Re: Airplane crashing into Atlanta-NAP
by owen＠DeLong.SJ.CA.US 27 Oct '96

27 Oct '96

> > 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. > > > Mae-East is pretty good example of this still.. those DC batteries didn't > appear to me to be featherweights, up ontop of concrete blocks. I don't recall > seeing DC batteries in mae-west up in the building.. where did they put > them? > They are indeed upstairs, one floor below the node. > The cost of building a datacenter in a multistory building is much much > greater than a one story building, as we have learned. The only benefit we > have seen from the multistories, is the usual coincidence that they are > more likely to be "lit" with glass, than a vacant one story.. unless you are > acquiring an old bell building or something.. owen ;) > Actually, it depends on the multi-story building. Of course, we did acquire an old Telco suite. :-) > rob > > Owen

1 0

Re: Provider credibility - does it matter? was Re: Inter-provider relations
by Vadim Antonov 27 Oct '96

27 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. That already was beaten to death. However, i repeat the argument: Big Provider Customer A ---[POP] ------------- 1000 miles -----------[POP] | IXP | Customer B ------[POP]-1 mile-[POP] Small Provider When customers A and B talk Big Provider pays to get them through 1000 miles. Small Provider pays for 1 mile. Note that i didn't even talk about less measurabe, but way too more important things like hosting of information suppliers. Say, Big Provider connects 1000 web sites; Small Provider hosts 1 site -- benefit from peering in terms of Web site diversity to the Big Provider's customers is 0.1%. To Small Provider's customers the benefit of peering is 99.9%. 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. --vadim

8 11