From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751170AbWGKSI3 (ORCPT ); Tue, 11 Jul 2006 14:08:29 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1751171AbWGKSI3 (ORCPT ); Tue, 11 Jul 2006 14:08:29 -0400 Received: from mail.atl.external.lmco.com ([192.35.37.50]:15504 "EHLO enterprise.atl.lmco.com") by vger.kernel.org with ESMTP id S1751170AbWGKSI2 (ORCPT ); Tue, 11 Jul 2006 14:08:28 -0400 Message-ID: <44B3E920.9070907@atl.lmco.com> Date: Tue, 11 Jul 2006 14:08:32 -0400 From: Jonathan Walsh User-Agent: Thunderbird 1.5.0.4 (X11/20060516) MIME-Version: 1.0 To: linux-kernel@vger.kernel.org Cc: "Gautam H. Thaker" , mingo@elte.hu Subject: RE: ~5x greater CPU load for a networked application when using 2.6.15-rt15-smp vs. 2.6.12-1.1390_FC4 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org As a follow up to previous emails (Gautam Thaker, Ingo Molnar, Ted Tso, et. al.) on the subject of large CPU overhead by the RT kernel when under heavy network load, I ran the following test in order to get more reasonable data. I have 19 nodes with 20 "virtual" node processes sending UDP messages to a single host at a rate of 100Hz for 38,000 packets per second. Using cyclesoak to determine cpu usage (over 240 samples, 1 sample per second), I found the following results: RT kernel: linux-2.6.17-rt1-uni Mean: 48.9% Variance: 5.91 Standard Deviation: 2.43 Standard kernel: Standard Fedora Core 4 Mean: 23.2% Variance: 0.237 Standard Deviation: 0.4867 Thus I found the average cpu load on the RT kernel to be 2.11 times that of the standard kernel. Hopefully this information will be of some use. -Jonathan Walsh Distributed Processing Lab; Lockheed Martin Adv. Tech. Labs