Date: Fri, 24 Apr 1998 11:02:22 -0700 From: Julian Elischer <julian@whistle.com> To: Luigi Rizzo <luigi@labinfo.iet.unipi.it> Cc: Kenjiro Cho <kjc@csl.sony.co.jp>, current@FreeBSD.ORG Subject: Re: Bandwidth throttling etc. Message-ID: <3540D3AE.52BFA1D7@whistle.com> References: <199804241155.NAA21152@labinfo.iet.unipi.it>
next in thread | previous in thread | raw e-mail | index | archive | help
Luigi Rizzo wrote: > > > > actually i was going to ask next if there are stats on the size of > packets, to see if it would be worthwhile increasing the size of an > MBUF to 256 bytes. > > With the increasing use of TCP and IP options, possibly longer > query strings to http, and large nowaday's memories, > it might be useful to improve performance. > > E.g. 20ms PCM audio packets are 160 bytes of data, plus headers etc, they > could fit there. Same for 80ms GSM frames... I notice that teh mpath module released the other day includes as part of it's patch, an upgrade of teh mbuf size to 256 bytes.. works for him so I guess it should work for us.. :-) Also, on the topic of BPF for ip filter descriptions. That is the obvious thing to investigate, but I'm not convinced that it's optimal. It is already resent I'll admin, but it does a lot of work to isolate fields etc, because it's protocol intependent, and each bytecode needs interpretation. It's not at all very efficient, even if it IS very general. Having spent a while looking at it yesterday, I think that our knowledge that the filter is working with an IP packet can be used to produce a much more efficient filter than BPF. As a 'language', ipfilter or ipfw would each be a better place to start. julian To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3540D3AE.52BFA1D7>