<snip> I've been thinking about other information that could be conveyed in communities. For instance, bandwidth, delay and packet loss. If each router along the way modifies such a community (should probably be an extended one) then a much richer set of information would be available to multihomers to aid in route selection. Hmmm well isn't this not what OSPF/EIGRP are *supposed* to do but they only calculate the values once. I guess the logic being you don't want SPF calcs killing all the routers in an area. For BGP you could either follow this logic or update the adverts every "x" minutes/hours, does this offer any more information given it is so old. Or you change the advert more dynamically and you send Cisco shares soaring as we need to upgrade routers :o). Mind you is this not a bit like QoS Policy Propagation in BGP in reverse (in so much as you are asking for delay/bw/packet loss guarantees)? I think even basic communities add value to a clued in peer/customer. Being able to limit the distance of your adverts for example. For content heavy sites this can be useful if you want to say peer with joe-bloggs-AS so that people in joe-bloggs city get "better" access. An example of this is ip-plus see the URL below: http://www.ip-plus.net/technical/config_bgp-en.html#5 Regards, Kevin
Date: Tue, 16 Oct 2001 09:55:12 +0100 From: Kevin Gannon <kgannon@lancomms.ie>
I've been thinking about other information that could be conveyed in communities. For instance, bandwidth, delay and packet loss. If each router along the way modifies such a community (should probably be an extended one) then a much richer set of information would be available to multihomers to aid in route selection.
[ snip ] I'd toyed with the bandwidth idea... let's say that we create 0xffff:1, :2, :3, and :4 to mean sub-T1, T3, OC3, and >= OC12. Arbitrary and simplified for sake of example. Whenever a router forwards, it tags with the appropriate speed. Thus, if :1 is set, you _know_ this path has a sub-T1 link. Of course, some places might use bogus tags to try hiding their connectivity... but an upstream could use a route-map to let the truth escape. Thus, with proper administration by upstreams, !set{:1|:2|:3|:4} would indicate no participants along the way. However, if this were to gain momentum among the biggest three dozen ASen, I think that we'd have a very accurate system. Even the top one dozen could do a very good job, considering how many connect directly to them... Eddy --------------------------------------------------------------------------- Brotsman & Dreger, Inc. - EverQuick Internet Division Phone: +1 (316) 794-8922 Wichita/(Inter)national Phone: +1 (785) 865-5885 Lawrence --------------------------------------------------------------------------- Date: Mon, 21 May 2001 11:23:58 +0000 (GMT) From: A Trap <blacklist@brics.com> To: blacklist@brics.com Subject: Please ignore this portion of my mail signature. These last few lines are a trap for address-harvesting spambots. Do NOT send mail to <blacklist@brics.com>, or you are likely to be blocked.
participants (2)
-
E.B. Dreger -
Kevin Gannon