From owner-freebsd-hackers Wed Jun 9 11:38:19 1999 Delivered-To: freebsd-hackers@freebsd.org Received: from abalaea.ircache.net (abalaea.scd.ucar.edu [128.117.28.9]) by hub.freebsd.org (Postfix) with ESMTP id 1A91015564 for ; Wed, 9 Jun 1999 11:38:16 -0700 (PDT) (envelope-from glenn@abalaea.ircache.net) Received: from localhost (glenn@localhost) by abalaea.ircache.net (8.9.2/8.9.1) with ESMTP id MAA20169; Wed, 9 Jun 1999 12:37:13 -0600 (MDT) (envelope-from glenn@abalaea.ircache.net) Date: Wed, 9 Jun 1999 12:37:13 -0600 (MDT) From: Glenn Chisholm To: Ted Faber Cc: Dan Moschuk , hackers@FreeBSD.ORG Subject: Re: hundreds of sockets stuck in TIME_WAIT In-Reply-To: <199906091755.KAA13333@boreas.isi.edu> Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: owner-freebsd-hackers@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG > And more directly for you, if they're not harming your performance, I > wouldn't do *anything*. They do serve a purpose. > When benchmarking proxies we find the MSL creates a resource problem for us with Web Polygraph (http://polygraph.ircache.net/). I must admit that we have never measured this as a performance issue. However I believe that Pei Cao's group has and found it to be one due to the way sockets were maintained in the kernel. I will look for a reference for their work. I think that one of her students submitted a patch to correct this issue in FreeBSD, so it may not be a problem in 3.1/3.2. If you do want to hack the MSL : /usr/src/sys/netinet/tcp_timer.h #define TCPTV_MSL ( 3*PR_SLOWHZ) /* max seg lifetime (hah!) */ That will make the MSL 3 seconds. glenn To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message