Date: Thu, 26 Jul 2001 18:05:01 -0700 (PDT) From: Hugh LaMaster <lamaster@nas.nasa.gov> To: freebsd-hackers@freebsd.org Subject: Re: Network performance tuning. Message-ID: <Pine.GSO.4.05.10107261724120.28652-100000@kinkajou.arc.nasa.gov> In-Reply-To: <3B53180D.F73197A2@mindspring.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Mon, 16 Jul 2001, Terry Lambert wrote: > Matt Dillon wrote: > > Also, the algorithm is less helpful when it has to figure out the > > optimal transmit buffer size for every new connection (consider a web > > server). I am considering ripping out the ssthresh junk from the stack, > > which does not work virtually at all, and using the route table's > > ssthresh field to set the initial buffer size for the algorithm. > > The "state of the art" for initial window size is flow monitoring > at the next hop router, with feedback to the host system. > > A gross approximation is called "fast start". ClickArray has a > patch for this, which could potentially be released back to > FreeBSD; I will check with TPTB... it's a pretty trivial version > of the RFC'ed initial window scaling. Two threads here, the first dealing with MBUFs, buffer allocation, &etc., which I'm not mentioning (though clearly there is useful work to do). The other thread is SACK, & other TCP performance improvements. The "state of the art" in this case may be "TCP Westwood": http://www.cs.ucla.edu/NRL/hpi/papers.html http://www.cs.ucla.edu/NRL/hpi/papers/2001-mobicom-0.ps.gz http://www-dee.poliba.it/dee-web/Personale/mascolo.html (articles on "Smith Principle", control theory applied to TCP) The mobicom paper "faster recovery" algorithm is, AFAIK, the state of the art, providing somewhat improved throughput on congested links, and, significantly improved throughput on lossy wireless links that many of us use as the last hop. -- Hugh LaMaster, M/S 233-21, Email: lamaster@nas.nasa.gov NASA Ames Research Center Or: lamaster@nren.nasa.gov Moffett Field, CA 94035-1000 Or: lamaster@kinkajou.arc.nasa.gov Phone: 650/604-1056 Disc: Unofficial, personal *opinion*. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.GSO.4.05.10107261724120.28652-100000>