From owner-freebsd-net@FreeBSD.ORG Sat Oct 30 17:41:53 2010 Return-Path: Delivered-To: freebsd-net@freebsd.org Received: from mx1.freebsd.org (mx1.freebsd.org [IPv6:2001:4f8:fff6::34]) by hub.freebsd.org (Postfix) with ESMTP id 1BA98106566C for ; Sat, 30 Oct 2010 17:41:53 +0000 (UTC) (envelope-from emaste@freebsd.org) Received: from mail1.sandvine.com (Mail1.sandvine.com [64.7.137.134]) by mx1.freebsd.org (Postfix) with ESMTP id CF0348FC1D for ; Sat, 30 Oct 2010 17:41:52 +0000 (UTC) Received: from labgw2.phaedrus.sandvine.com (192.168.222.22) by WTL-EXCH-1.sandvine.com (192.168.196.31) with Microsoft SMTP Server id 14.0.694.0; Sat, 30 Oct 2010 13:41:52 -0400 Received: by labgw2.phaedrus.sandvine.com (Postfix, from userid 10332) id 1E66433C00; Sat, 30 Oct 2010 13:41:52 -0400 (EDT) Date: Sat, 30 Oct 2010 13:41:52 -0400 From: Ed Maste To: Paul Thornton Message-ID: <20101030174152.GB41042@sandvine.com> References: <1519248747.20101028232111@yandex.ru> <1452146D-A590-4676-A662-14D0EEE82152@mac.com> <606859717.20101029093926@yandex.ru> <4CCC0243.8060507@prt.org> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Disposition: inline In-Reply-To: <4CCC0243.8060507@prt.org> User-Agent: Mutt/1.4.2.1i Cc: freebsd-net@freebsd.org Subject: Re: Polling slows down bandwidth X-BeenThere: freebsd-net@freebsd.org X-Mailman-Version: 2.1.5 Precedence: list List-Id: Networking and TCP/IP with FreeBSD List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Sat, 30 Oct 2010 17:41:53 -0000 On Sat, Oct 30, 2010 at 12:32:19PM +0100, Paul Thornton wrote: > I've been doing testing with FreeBSD 8 and em interfaces recently, and > my experience agrees with Chuck's statement - that polling makes things > worse when you use new (anything in the last 2 or 3 years) hardware with > good quality gigabit ethernet interfaces. There are some deficiencies in the current polling algorithm that will cause it to perform less than optimally (it will temporarily stop processing packets even though it is consuming less CPU than requested). I have some changes that I plan to bring into the tree to improve this situation. For recent high quality hardware though I expect you'll get roughly equivalent performance from polling and standard opteration. -Ed