- NANOG - lists.nanog.org

Re: Exchanges that matter...
by Dave O'Shea 06 Dec '96

06 Dec '96

> > Do you see symmetrical flows in and out of the country or do you > > see more inbound that outbound? > > There is a third possibility, of course. In our case we see more > outbound traffic towards the USA then inbound traffic from the USA, > because our UK customers have Web sites popular in the USA. .. So we have not only a trade deficit, but an "ack" deficit to keep politicians awake at night.. :-)

1 0

February meeting
by Susan R. Harris 06 Dec '96

06 Dec '96

Greetings - Here's some preliminary information about the upcoming NANOG meeting, to be held Monday-Tuesday, February 10-11, at the Grand Hyatt in San Francisco. - The meeting will last a bit longer this time - we'll run until 4:00 p.m. on Tuesday the 11th, rather than ending at noon as in the past. We'll send out a list of preliminary topics before the holidays, and will welcome your suggestions for further agenda items. The agenda will be posted at: http://www.nanog.org/ - There will be a NetSCARF tutorial on Sunday, February 9, from 1:00- 5:00 p.m. That evening, we'll be offering a tutorial session for new ISPs (topic to be announced). - Registration begins January 2. To help us with record-keeping, everyone will be asked to register via the Web. You'll have several options for paying the $100 attendance fee, depending on whether or not you want to submit payment electronically: - Fill out the registration form and send it to us over the Web with your credit card info. - Fill out the registration form and send it to us over the Web. Call and give us your credit card info over the phone. - Fill out the registration form and send it to us over the Web. Then fax or mail us your credit card info or mail us a check for the registration fee. We'll send out further details in a later message. Looking forward to seeing you at the meeting!

2 1

Re: Exchanges that matter...
by Vadim Antonov 06 Dec '96

06 Dec '96

salo(a)msc.edu (Tim Salo) wrote: >> FDDI is not a WAN technology. Losing 20% of bandwidth of a 10 ft >> piece of fiber is one thing. Losing 20% of bandwidth of a >> $3M/yr circuit is quite different. >I don't understand. Are you trying to say that price/performance >analysis is appropriate for LAN technologies, but not for WAN technologies? Quite opposite. Price of "transmission" in LANs is negligible as compared to the price of interface cards and switches. So a good price/performance of the overall system can be achieved by simplifying those switches or interfaces, particularly by using less efficient methods of transmission. The price of transmission facility in WANs easily exceed price of switching equipment by an order of magnitude or more. Therefore it is very silly to sacrifice transmission capacity for the sake of easier switching in WANs. >Now, I could understand if you said, "I analyzed ATM WAN services for >one configuration once and concluded that other technologies provided >better cost/performance in all possible configurations," or even >"I don't think ATM products and standards are mature enough for my >application." I don't think it will work as anything but sync muxes with silly framing in the real life WANs. I published my reasons for thinking so. >However, you seem to be saying that wide-area technologies >should be assessed based on total cost and overhead, rather than on >cost/performance... No, i simply left out a simple conjecture. --vadim

1 0

Re: Exchanges that matter...
by salo＠msc.edu 06 Dec '96

06 Dec '96

> Date: Thu, 5 Dec 1996 14:56:00 -0800 > From: Vadim Antonov <avg(a)pluris.com> > To: paul(a)vix.com, salo(a)msc.edu > Subject: Re: Exchanges that matter... > Cc: nanog(a)merit.edu [slashing and reordering your response slightly] > ATM as a LAN is probably ok, if it can compete with Gb Ethernet > price/performance-wise. Precisely my point, (which you managed to delete in your response). Price/performance seems like a pretty good analysis technique, at least for technologies that pass some basic hurdles, like "working." > FDDI is not a WAN technology. Losing 20% of bandwidth of a 10 ft > piece of fiber is one thing. Losing 20% of bandwidth of a > $3M/yr circuit is quite different. I don't understand. Are you trying to say that price/performance analysis is appropriate for LAN technologies, but not for WAN technologies? Now, I could understand if you said, "I analyzed ATM WAN services for one configuration once and concluded that other technologies provided better cost/performance in all possible configurations," or even "I don't think ATM products and standards are mature enough for my application." However, you seem to be saying that wide-area technologies should be assessed based on total cost and overhead, rather than on cost/performance... -tjs

1 0

ISPACs
by Tony Li 06 Dec '96

06 Dec '96

Folks, I'd like to point your attention to the following ID. I would appreciate any comments. For the lack of a better place, I'd ask that discussion be on the cidrd mailing list. Thanks, Tony A New Internet-Draft is available from the on-line Internet-Drafts directories. Title : Internet Service Provider Address Coalitions (ISPACs) Author(s) : T. Li Filename : draft-li-ispac-00.txt Pages : 5 Date : 11/25/1996

2 4

Re: Exchanges that matter...
by Vadim Antonov 06 Dec '96

06 Dec '96

Bill Sommerfeld <sommerfeld(a)apollo.hp.com> wrote: As is, practically all Internet traffic is long-haul. >Folks doing part-time telecommuting are likely to generate traffic >with very high geographic locality; unfortunately, in a lot of cases, >because of routing nonoptimalities, they wind up generating >"long-haul" traffic instead.. Yes, although most employers either provide their private dial-in facilities or get a good "package" deal from a local ISP on behalf of its employees. Still, Web constitutes the bulk of traffic... I do not argue that local exchanges between regionals are entirely worthless; it's that they won't have large impact on what happens at "large" exchange points (or private interconnects). --vadim

1 0

Re: The Cidr Report
by Paul Ferguson 06 Dec '96

06 Dec '96

At 01:20 PM 12/5/96 -0500, Curtis Villamizar wrote: >> >> >Date Prefixes >> >231196 40687 >> >241196 40908 >> >251196 40844 >> >261196 41043 >> >271196 41074 >> >281196 39803 >> >291196 39735 >> > >People withdraw their prefix announcement due to the holidays? > Perhaps coincidence. Perhaps not. :-) Date Prefixes 281196 39803 291196 39735 301196 39676 011296 39610 021296 39948 031296 40057 041296 40242 051296 41491 - paul

1 0

Re: Exchanges that matter...
by Bill Sommerfeld 06 Dec '96

06 Dec '96

As is, practically all Internet traffic is long-haul. I can believe that this likely to be true for web browsing, but I *know* that at least one Boston-area ISP has a "private interconnect" with MIT because they had enough traffic locality to warrant it. Folks doing part-time telecommuting are likely to generate traffic with very high geographic locality; unfortunately, in a lot of cases, because of routing nonoptimalities, they wind up generating "long-haul" traffic instead.. - Bill

1 0

Re: Exchanges that matter...
by Vadim Antonov 06 Dec '96

06 Dec '96

Wayne Bouchard <web(a)typo.org> wrote: >Actually... many of the private exchanges are there for several other >reasons as well. First, it can be convenient to set up a connection >with a net you're passing lots of traffic to so you can go direct to >them instead of through 7 or 8 odd routers outside your >net. (obviously) We were talking about tier 1 exchanges (i.e. between nation-wide and world-wide backbones). They by definition pass a lot of traffic between each other. >Second, it can be significantly cheaper to split the costs between the >participants and not have to pay a third party for access to the >exchange. (heck of a lot easier too..) Well, if you have 6 providers you'll need 15 private two-party exchanges to accomodate full connectivity in a region, vs a single public facility. It is a classical example of economies of scale. Consolidation of exchange facilities makes things a lot cheaper. (Providing there's a way to sustain that much traffic in one place). >Third, and going along with the first point, private exchanges can >help to more geographically orient your local corner of the >network. That way, to get across town, you don't have to go to >mae-west, over to denver, and back down to phoenix to travel 6 miles. > >This also serves to take a little traffic AWAY from the public NAPs >and can help reduce congestion and problems there. That would be fine, if traffic had high locality. As is, practically all Internet traffic is long-haul. I.e. i strongly doubt benefits of micro-exchanges will outweigh their drawbacks. >The principle disadvantage of lots of small exchange points, as far as >I see it, is that there may well be potential for A) many more paths >to get from A to B, meaning that each router will hold that many more >BGP entries and B) possibly a greater potential for route flap. C) a lot more facilities not generating revenue. --vadim

1 0

RE: Exchanges that matter...
by Danny Stroud 06 Dec '96

06 Dec '96

Thanks, this is fundamentally the argument I was trying to make. Obviously, Wayne was much more articulate and does a better job. Also, sorry about the confusion on my email address. This is my personal box. See below for my true identity. des ----------------------- Danny E. Stroud WNA, Inc. 9250 East Costilla Avenue, Suite 400 Englewood, CO 80112 dstroud(a)wna.net office: 303-645-1904 fax: 303-792-3869 cellular: 303-903-0911 ---------- From: owner-nanog(a)merit.edu on behalf of Wayne Bouchard Sent: Thursday, December 05, 1996 4:39 PM To: Vadim Antonov Cc: Danny Stroud; nanog(a)merit.edu Subject: Re: Exchanges that matter... > Danny Stroud <dannystroud(a)msn.com> wrote: > > >So how do you reconcile the increasing number of private facilities? Although > >not exchange points in the truest sense ala MAEs and NAPs, they do carry an > >increasing percentage of cross-network traffic. The number of these types of > >facilities will grow. Maybe our difference here is that I count these and > >others do not. des > > Er. Private exchanges are there only because public exchanges are running > out of gas. They have the same problems with routing as public exchanges do. Actually... many of the private exchanges are there for several other reasons as well. First, it can be convenient to set up a connection with a net you're passing lots of traffic to so you can go direct to them instead of through 7 or 8 odd routers outside your net. (obviously) Second, it can be significantly cheaper to split the costs between the participants and not have to pay a third party for access to the exchange. (heck of a lot easier too..) Third, and going along with the first point, private exchanges can help to more geographically orient your local corner of the network. That way, to get across town, you don't have to go to mae-west, over to denver, and back down to phoenix to travel 6 miles. This also serves to take a little traffic AWAY from the public NAPs and can help reduce congestion and problems there. The principle disadvantage of lots of small exchange points, as far as I see it, is that there may well be potential for A) many more paths to get from A to B, meaning that each router will hold that many more BGP entries and B) possibly a greater potential for route flap. -Wayne

1 0