- NANOG - lists.nanog.org

PSI statement on C&W peering issue
by Sean Donelan 05 Jun '01

05 Jun '01

PSI statement from their website Cable & Wireless chose to terminate connectivity with PSINet on 2June01. Over 90% of the traffic that used to be routed through C&W is now being routed via other means through our robust global free peering infrastructure. The remaining 10% or so represents C&W customers that have been deliberately cut off from PSINet by C&W. While PSINet is ready and willing to re-establish connectivity with C&W at any time, it is up to C&W to choose to reverse their previous decision. In the meantime, PSINet can offer services directly to those C&W customers that are affected. http://www.psi.net/cgi-bin/netstatus.pl5 Does the 10% number mean C&W's network reach has shrunk, or that a growing number of networks are multihomed?

4 3

RE: 95th Percentile = Lame
by Andrew Odlyzko 05 Jun '01

05 Jun '01

The discussion over the last few days has been fascinating. I would like to inject a few remarks. The main one is that there is a huge literature on pricing of communication services and other utilities. The problem is that statements such as > Ideally, the price should match as closely as possible the actual cost to > provide the service. are very appealing, and even agree with conventional economic doctrine, but founder on the difficulty of determining what "the actual cost to provide the service" is. In addition, there are other conventional economic arguments that say prices should match not costs, but rather willingness to pay. Electricity pricing, which has been mentioned here several times, is an interesting case. Original pricing was flat rate (so many dollars or cents for each lightbulb for each month). That was discarded very quickly, largely because of a crucial factor that distinguishes electricity production from telecommunications, namely high marginal costs. Except for hydro (and to some extent nuclear) power, there is a measureable and substantial cost in paying for the fuel that provides each kilowatt-hour of "juice." Even that, though, does not deal with the issues of fixed costs (the generating plant and the transmission lines). How to allocate those costs to various consumers led to extensive discussions and experimentation about a century ago, far more sophisticated than anything that was done in telecommunications at that time or even now. What we have today is usually a combination of fixed rates (dependent on capacity of link) and usage charges (straight fees per kilowatt-hour). Experiments with time-of-day pricing have had mixed outcomes, and there is little of it going on. This might change with smarter appliances, but then it might not. The Internet does not have the high marginal costs that electricity involves. Hence the economic case is different, but even so, theory does not provide an unambiguous answer as to what the answer is. I have several papers that discuss pricing of Internet and other communications (and even more general) services. As a result of the investigations described in those papers, I do come down on a particular side of the debate (namely in favor of flat rates), but also provide extensive references for other arguments. There is a short, 6-page extended abstract entitled "Internet pricing in light of the history of communication," <http://www.research.att.com/~amo/doc/history.communications1.pdf>, (to appear in Proc. ITCom 2001), the full 40-page paper "Internet pricing and the history of communications," <http://www.research.att.com/~amo/doc/history.communications1b.pdf>, to appear in "Computer Networks," and a longer yet and more detailed 160-page manuscript "The history of communications and its implications for the Internet," <http://www.research.att.com/~amo/doc/history.communications0.pdf>. (Replace .pdf in the URLs above with .ps if you are a fan of PostScript.) These might provide some amusement and possibly even enlightenment. Andrew Odlyzko ************************************************************************ Andrew Odlyzko amo(a)research.att.com AT&T Labs - Research voice: 973-360-8410 http://www.research.att.com/~amo fax: 973-360-8178 ************************************************************************

1 0

Agis, PSI, next C&W?
by Sean Donelan 05 Jun '01

05 Jun '01

On Mon, 04 June 2001, "Vivien M." wrote: > It's not likely to be a big technical deal, but the irony I find to be quite > prominent. First PSI forced others to make transit arrangements because of > their greed, and now CW is possibly making PSI do the same, for probably the > same motives. The funny thing about "peering policies" is they are no set policy. Networks can write whatever requirements or non-requirements they choose. Claims that one network or another doesn't meet arbitrary "peering requirements" should always be examined skeptically. Who set these requirements? If you put all the peering requirements into a barral and drew them out randomly, would provider P or C still met any arbitrary requirements from the barrel? Some regular NANOG attendees remember when AGIS implemented their new peering policy. Now, nobody cares about AGIS's peeering requirements. Some regular NANOG attendees remember when PSI implemented their new peering policy. Now, nobody cares about PSI's peering requirements. Cable & Wireless has implemented their new peering policy. I wonder how long it will be before nobody cares about C&W's peering requirements. After their acquisition of InternetMCI customers, C&W was extremely slow in upgrading their network and even slower in upgrading their peering connections. PSI concentrating in international expansion, so their US network is probably smaller than some other US-based providers. On their other hand, if you evaluated PSI as an international provider you might find they meet the "peering requirements" for international providers since other US-based providers are relatively weak elsewhere in the world. Should C&W purchase international transit from PSI? If C&W doesn't want to peer with PSI's international network, I guess C&W will have to decide if they want to buy transit or do without.

1 0

A moment of silence.
by Leo Bicknell 05 Jun '01

05 Jun '01

router>show int fddi5/0/0 Fddi5/0/0 is down, line protocol is down Description: FDDI to mae-east Jun 1 14:36:38.190: %LINK-3-UPDOWN: Interface Fddi5/0/0, changed state to down Jun 1 14:36:39.190: %LINEPROTO-5-UPDOWN: Line protocol on Interface Fddi5/0/0, changed state to down It appears that MFS finally went and did it, and the mae-east we all came to know and love has finally been turned off. Please bow your heads for a moment of silence. -- Leo Bicknell - bicknell(a)ufp.org Systems Engineer - Internetworking Engineer - CCIE 3440 Read TMBG List - tmbg-list-request(a)tmbg.org, www.tmbg.org

11 13

Re: PSI statement on C&W peering issue
by Richard A. Steenbergen 04 Jun '01

04 Jun '01

On Mon, Jun 04, 2001 at 04:32:45PM -0700, Sean Donelan wrote: > > Does the 10% number mean C&W's network reach has shrunk, or that a > growing number of networks are multihomed? I'm gonna go with "PSI made up a number that sounded good". -- Richard A Steenbergen <ras(a)e-gerbil.net> http://www.e-gerbil.net/ras PGP Key ID: 0x138EA177 (67 29 D7 BC E8 18 3E DA B2 46 B3 D8 14 36 FE B6)

1 0

Re: C&W Peering
by Patrick W. Gilmore 04 Jun '01

04 Jun '01

At 05:44 PM 6/4/2001 -0400, Travis Pugh wrote: >route-views.oregon-ix.net concurs: > >route-views.oregon-ix.net>sh ip bgp regex ^3561_174_ > >route-views.oregon-ix.net> Yes, but if one or the other has backup peering, it would not look like that. It would look like _3561_.*_174_ or _174_.*_3561_ - prolly the former since AS3561 gives route-views a feed, but AS714 does not. Looking in route-views for those two patterns, I see only a few routes under 3561.*_174_, probably leakage. There are no routes of the form _174_.*_3561_. Since route-views does have a feed from AS3561, I would say it is official. Cable and Wireless cannot reach PSI.net. Congratulations ladies & gentlemen. The first intentional, prolonged, significant (for some values of "significant" :) outage on the Internet. And we were all here to see it.... Wow, in the same week MAE-East dies. Sad time for the 'Net. :(( >-travis -- TTFN, patrick

5 5

Traceloop between PSI and C&W
by Steve Schaefer 04 Jun '01

04 Jun '01

Here's a quick link so that anyone can check to see actual traces between C&W and PSI. When (if?) they find each other again, you'll be able to see the actual routes (in both directions): http://www.traceloop.com/cgi-bin/traceloop.cgi?sid=fnryumci&dn=AS3561.a1 This is a special link -- you'll have to log out and log in if you want to use the normal service. Every Traceloop is being done by the same two test points, so there could be some blocking when this post first hits... Hope that helps. Steve Schaefer Dashbit - The Leader In Internet Topology www.dashbit.com www.traceloop.com

1 0

Re: C&W Peering
by Kevin Loch 04 Jun '01

04 Jun '01

This is sure to create a number of "big pipe orphans". I wonder if this will create a surge in multihoming attempts? How could you discourage that now? Unlike the recent DSL disasters, you can't just say "buy a T1 if you want reliable service". Even if you are not a PSI customer, it would be foolish not to multihome now. Did C&W consider the route table effects of this new routing policy? KL "Patrick W. Gilmore" wrote: > > At 05:44 PM 6/4/2001 -0400, Travis Pugh wrote: > > >route-views.oregon-ix.net concurs: > > > >route-views.oregon-ix.net>sh ip bgp regex ^3561_174_ > > > >route-views.oregon-ix.net> > > Yes, but if one or the other has backup peering, it would not look like > that. It would look like _3561_.*_174_ or _174_.*_3561_ - prolly the > former since AS3561 gives route-views a feed, but AS714 does not. > > Looking in route-views for those two patterns, I see only a few routes > under 3561.*_174_, probably leakage. There are no routes of the form > _174_.*_3561_. > > Since route-views does have a feed from AS3561, I would say it is > official. Cable and Wireless cannot reach PSI.net. > > Congratulations ladies & gentlemen. The first intentional, prolonged, > significant (for some values of "significant" :) outage on the > Internet. And we were all here to see it.... > > Wow, in the same week MAE-East dies. Sad time for the 'Net. :(( > > >-travis > > -- > TTFN, > patrick

3 2

Re: C&W Peering
by Sean Donelan 04 Jun '01

04 Jun '01

On Mon, 04 June 2001, Simon Lockhart wrote: > Well, I assumed PSI and/or C&W would have had some backup transit, if they > were going to play that game, but... > By defintion a "tier 1" provider does not have any transit, backup or otherwise. If tier 1 providers terminate peering, there are no alternate routes between them. However, customers who are multihomed may be able to reach both of them. Likewise customers of other peers can reach both of them. But customers exclusively connected to one or the other battling peer can't reach customers exclusively attached to the other. Its been a while since we've had a real "tier 1" peering battle. Last time ANS lasted a couple of weeks before they caved in and joined the CIX after the CIX filtered out non-member routes.

1 0

Re: 95th Percentile again (was RE: C&W Peering Problem?)
by Richard A. Steenbergen 04 Jun '01

04 Jun '01

On Sun, Jun 03, 2001 at 10:48:11PM -0400, Jim Mercer wrote: > > On Sun, Jun 03, 2001 at 01:34:08PM -0400, Alex Rubenstein wrote: > > > most of my customers are ethernet connected, and some customers share > > > an interface. i got into this before it was cheap to do 802.11q > > > switching, so my billing system needed to deal with multiple customers > > > on a single ethernet. > > > > Thats irrelevant; look at the counters on the customers switch-port. > > Doesn't matter what VLAN (or none) they are on. > > ah, well, again, i was using multiport bsd implementations, not > switches. > > as it stands, i'm fairly content using the cache flow from the > cisco's. > > one day, i might actually install a full 802.11q compliant switching > framework, but, well, not today. Come on, even D-Link is making switches with 802.1q... http://www.dlink.com/products/switches/des6000/ Go search your couch for loose change. -- Richard A Steenbergen <ras(a)e-gerbil.net> http://www.e-gerbil.net/ras PGP Key ID: 0x138EA177 (67 29 D7 BC E8 18 3E DA B2 46 B3 D8 14 36 FE B6)

1 0