From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1767016AbXDEOss (ORCPT ); Thu, 5 Apr 2007 10:48:48 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1767025AbXDEOss (ORCPT ); Thu, 5 Apr 2007 10:48:48 -0400 Received: from stinky.trash.net ([213.144.137.162]:52602 "EHLO stinky.trash.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1767016AbXDEOsr (ORCPT ); Thu, 5 Apr 2007 10:48:47 -0400 Message-ID: <46150C49.4050800@trash.net> Date: Thu, 05 Apr 2007 16:48:41 +0200 From: Patrick McHardy User-Agent: Debian Thunderbird 1.0.7 (X11/20051019) X-Accept-Language: en-us, en MIME-Version: 1.0 To: leon zadorin CC: linux-kernel@vger.kernel.org Subject: Re: "no hz"/"dynamic tick" option and TC/TBF (token bucket filter) behaviour... References: <26d2cb010704050736y775b6b09g67fd40344cd2208d@mail.gmail.com> In-Reply-To: <26d2cb010704050736y775b6b09g67fd40344cd2208d@mail.gmail.com> X-Enigmail-Version: 0.93.0.0 Content-Type: text/plain; charset=ISO-8859-15 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org leon zadorin wrote: > 2) for kernels which do have the aforementioned option enabled: how > does this affect the "tc" (traffic control) utility and the Token > Bucket Filter (tbf) queueing discipline (traffic-smoothing) > behhaviour? > > For example, the TBF doco suggests that due to the whole fixed Hz > thing in the kernel (e.g. 100Hz) the maximum rate of the TBF would > actually be limited to say 1mbit on Intel arch (with the default > mtu... but if mtu is incleased, so is the burstiness of the traffic > which is what TBF is trying to prevent in my deployment scenarios). > ... > > So what about the kernels which use this "dyna tick" option - in such > a kernel, will there be an interrupt generated (e.g. at microsec. > precision on most platforms) for every packet's send time (e.g. say at > TBF's rate is at 25Mbps and one sends out 1KB data packets... will > there be an interrput generated every... 300 microseconds?)... No, it doesn't affect the timer resolution. net-2.6.22 includes patches that switch all qdiscs to hrtimers, which allows to use much smaller burst rates.