From owner-freebsd-net Tue Aug 7 9:47:45 2001 Delivered-To: freebsd-net@freebsd.org Received: from panzer.kdm.org (panzer.kdm.org [216.160.178.169]) by hub.freebsd.org (Postfix) with ESMTP id A2CF537B405 for ; Tue, 7 Aug 2001 09:47:41 -0700 (PDT) (envelope-from ken@panzer.kdm.org) Received: (from ken@localhost) by panzer.kdm.org (8.9.3/8.9.1) id KAA46080; Tue, 7 Aug 2001 10:47:35 -0600 (MDT) (envelope-from ken) Date: Tue, 7 Aug 2001 10:47:34 -0600 From: "Kenneth D. Merry" To: stanislav shalunov Cc: net@FreeBSD.ORG Subject: Re: TCP problems with large window sizes on FreeBSD (GigaTCP) Message-ID: <20010807104734.A46031@panzer.kdm.org> References: <200108020503.f7253jY08157@prism.flugsvamp.com> <87y9p2qxvc.fsf@cain.internet2.edu> <87n15inp9r.fsf@cain.internet2.edu> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline User-Agent: Mutt/1.2i In-Reply-To: <87n15inp9r.fsf@cain.internet2.edu>; from shalunov@internet2.edu on Thu, Aug 02, 2001 at 05:08:32PM -0400 Sender: owner-freebsd-net@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.org On Thu, Aug 02, 2001 at 17:08:32 -0400, stanislav shalunov wrote: > stanislav shalunov writes: > > > I'll try increasing the value of TI_JSLOTS to 8192 (twice the number > > of 4K packets in 16MB window) and see if it makes a difference. > > FWIW, this seems to have improved the situation and the burst loss > doesn't happen anymore. For those who might care to look at the new > numbers, see http://www.internet2.edu/~shalunov/gigatcp/ > > There seems to be a problem with the ti driver in that it doesn't > recover from the situation of jumbo slots shortage right away after > supply of packets is removed, as one would expect it to. The > condition lingers for a long time afterwards. It seems to recover > eventually, but after quite different time for diffirent conditions: > sometimes it's seconds, sometimes it's minutes. The "slow" case > appears to be an example of fast periodic recovery, while in the > "stalled" case the recovery doesn't happen until the TCP connection > times out. The zero copy version of the ti(4) driver has a different jumbo buffer allocator that may reduce the intstance of problems. Ken -- Kenneth Merry ken@kdm.org To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-net" in the body of the message