From owner-freebsd-atm Thu Mar 7 7:18: 2 2002 Delivered-To: freebsd-atm@freebsd.org Received: from web.cs.ndsu.nodak.edu (web.cs.ndsu.NoDak.edu [134.129.125.7]) by hub.freebsd.org (Postfix) with ESMTP id 18C9137B492 for ; Thu, 7 Mar 2002 07:17:55 -0800 (PST) Received: (from tinguely@localhost) by web.cs.ndsu.nodak.edu (8.11.4/8.11.4) id g27FFDu63200; Thu, 7 Mar 2002 09:15:13 -0600 (CST) (envelope-from tinguely) Date: Thu, 7 Mar 2002 09:15:13 -0600 (CST) From: mark tinguely Message-Id: <200203071515.g27FFDu63200@web.cs.ndsu.nodak.edu> To: atm_roberto@yahoo.com, freebsd-atm@FreeBSD.ORG Subject: Re: (FreeBSD TCP) vs (Linux TCP) Sender: owner-freebsd-atm@FreeBSD.ORG Precedence: bulk List-ID: List-Archive: (Web Archive) List-Help: (List Instructions) List-Subscribe: List-Unsubscribe: X-Loop: FreeBSD.org on Thu, 7 Mar 2002 02:48:52 -0800 (PST), Russo Roberto said: > I am working to evaluate performance of > throughput between two PC connected between > a Fore-ATM card adapter to a network: > the netwotk is composed of a PVC of 4Mbps between > three routers and I am just looking to use > differente queuing discipline like FIFO-PQ-WFQ. > > I have configured the two PC with the same Operating > Systems: a Linux (Red-Hat) and a FreeBSD v4.3 (Kame > Patch) and I am using NETPERF to make throughput > test. I would suggest that you run FreeBSD 4.5. There was a long-hidden socket bug that under certain situation caused the socket mechanism to go idle for a lengthy period of time and this caused performance issues. This socket problem was fixed in the FreeBSD 4.5 release. You can read all about the problem in the freebsd-hackers mailing list under the topic of network or NFS performance bug in the early December 2001 time frame.. I cannot guarrentee that this is the very same thing that you are seeing, but without any other detail, it sounds simular. --mark tinguely. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-atm" in the body of the message