From owner-freebsd-questions Thu May 31 14: 3:39 2001 Delivered-To: freebsd-questions@freebsd.org Received: from root.com (root.com [209.102.106.178]) by hub.freebsd.org (Postfix) with ESMTP id 6793A37B422 for ; Thu, 31 May 2001 14:03:35 -0700 (PDT) (envelope-from dg@root.com) Received: (from dg@localhost) by root.com (8.11.2/8.11.2) id f4VKuWR85393; Thu, 31 May 2001 13:56:32 -0700 (PDT) (envelope-from dg) Date: Thu, 31 May 2001 13:56:32 -0700 From: David Greenman To: Bill Moran Cc: freebsd-questions@freebsd.org Subject: Re: delayed_ack Message-ID: <20010531135632.E25389@nexus.root.com> References: <3B16AFDD.F9167D5D@iowna.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <3B16AFDD.F9167D5D@iowna.com>; from wmoran@iowna.com on Thu, May 31, 2001 at 04:55:57PM -0400 Sender: owner-freebsd-questions@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.ORG >Hi all, > >I'm curious as to some particulars on this sysctl value. >(net.inet.tcp.delayed_ack) I recently found that setting it to 0 >produced a very noticable improvement in the performance of SMB traffic >from windows machine to a Samba server. This machine is 4.2-STABLE >cvsuped on April 5, 12:06. > >I've seen various messages in the mail archives with relation to this >setting and improvements gained. You can add me as one who saw such an >improvement. Specifically, on occasion, transfer operations would pause >(as if waiting for something to happen ... no network activity) and >after 45 seconds, complete. (This is using Windows 95 to connect to the >above described machine) Setting delayed_ack=0 eliminated this delay >completely. > >So ... here is my question: >Is there a problem with the delayed_ack code in FreeBSD? Is it Windows >that is the problem? It seems to me that there's something wrong with >SMB, since HTTP and ssh from Window's boxes works fine. This wasn't a >little performance improvement, this had the feel of something broken >getting fixed. We're talking 45 seconds, with zero network activity, >before the operation completed. > >Not a critical question, mostly curiousity. I think Windows is broken, but SMB should probably set TCP_NODELAY on the socket in any case. Setting delayed_ack=0 will make all TCP connections in the system essentially have the effect of setting TCP_NODELAY on them. -DG David Greenman Co-founder, The FreeBSD Project - http://www.freebsd.org President, TeraSolutions, Inc. - http://www.terasolutions.com Pave the road of life with opportunities. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message