Hi All, While I haven't read Bob's draft, so my comments may be naive, but RFC 2991 "Multipath Issues in Unicast and Multicast Next-Hop Selection" seems to recommend against round robin loadsharing, and arguably recommend against per packet load sharing at all.
While the RFC is about routers performing the loadsharing over multiple paths, I would think the issues it raises would probably also apply to hosts. Regards, Mark. > -----Original Message----- > From: Bob Hinden [mailto:[EMAIL PROTECTED]] > Sent: Thursday, 15 November 2001 11:05 > To: Christian Huitema > Cc: [EMAIL PROTECTED] > Subject: RE: draft-hinden-ipv6-host-load-sharing-00.txt > > > Christian, > > >Even if we wanted to somehow mandate load sharing, there are > generally > >issues with mandating a round robin approach, or in fact any specific > >algorithm. Round robin has two drawbacks: it hypothesizes that all > >routers are equal, which is very often not the case, and it > implies some > >explicit ordering of the requests, which can lead to > synchronizations. > >The all routers equal hypothesis is fine for dumb hosts, but > smart hosts > >can acquire a knowledge that this or that router usually performs > >better, which is enough to justify a "SHOULD" instead of > "MUST". The way > >to avoid synchronization is normally randomization, i.e. > pick routers in > >a random order rather than doing round robin. > > The issue of all routers not being equal is a generic > neighbor discovery > issue. Currently a host will add to it's default router list > any router it > receives routing advertisements. Rich Draves "Default Router > Preferences > and More-Specific Routers" draft addresses this issue and > allows a ranking > of the default routers. > > I think load sharing has value when there are multiple (equal > preference) > default routers. It is very common in the IPv4 world for > people to set up > parallel paths with two or more routers. In this case it is > useful for > hosts to spread their traffic among these routers. > > I will revise the draft to change from round-robin to random > order to avoid > synchronization, and also change it from a MUST to a SHOULD > to allow for > hosts with more information. > > Also, it may make sense to later combine this with Rich's > draft as they are > both dealing with related default router selection issues. > In this case > the load sharing would be done when there are multiple > default routers with > equal preference. > > Bob > > -------------------------------------------------------------------- > IETF IPng Working Group Mailing List > IPng Home Page: http://playground.sun.com/ipng > FTP archive: ftp://playground.sun.com/pub/ipng > Direct all administrative requests to [EMAIL PROTECTED] > -------------------------------------------------------------------- > -------------------------------------------------------------------- IETF IPng Working Group Mailing List IPng Home Page: http://playground.sun.com/ipng FTP archive: ftp://playground.sun.com/pub/ipng Direct all administrative requests to [EMAIL PROTECTED] --------------------------------------------------------------------
