From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from relay.sgi.com (relay1.corp.sgi.com [137.38.102.111]) by oss.sgi.com (Postfix) with ESMTP id B61BD7F4C for ; Mon, 18 Feb 2013 08:17:36 -0600 (CST) Received: from cuda.sgi.com (cuda3.sgi.com [192.48.176.15]) by relay1.corp.sgi.com (Postfix) with ESMTP id 581B18F8035 for ; Mon, 18 Feb 2013 06:17:36 -0800 (PST) Received: from mail-qc0-f176.google.com (mail-qc0-f176.google.com [209.85.216.176]) by cuda.sgi.com with ESMTP id ZvYQ9Kjzxz9MbqhO (version=TLSv1 cipher=RC4-SHA bits=128 verify=NO) for ; Mon, 18 Feb 2013 06:17:35 -0800 (PST) Received: by mail-qc0-f176.google.com with SMTP id n41so2081240qco.35 for ; Mon, 18 Feb 2013 06:17:35 -0800 (PST) MIME-Version: 1.0 In-Reply-To: <20130217212536.GY26694@dastard> References: <20130217212536.GY26694@dastard> Date: Mon, 18 Feb 2013 19:47:34 +0530 Message-ID: Subject: Re: High %user CPU in XFS From: Subranshu Patel List-Id: XFS Filesystem from SGI List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Errors-To: xfs-bounces@oss.sgi.com Sender: xfs-bounces@oss.sgi.com To: Dave Chinner Cc: xfs@oss.sgi.com > Page faults, CPU cache misses. XFS has a significantly large icache > and dcache footprint than ext4, so those are the likely causes I have a query. Will page fault, cache misses etc will increase the %user CPU utilization or %system CPU utilization. I thought it will affect the %system and not %user CPU usage? %user - Percentage of CPU utilization that occurred while executing at the user level (application). %system - Percentage of CPU utilization that occurred while executing at the system level (kernel). On Mon, Feb 18, 2013 at 2:55 AM, Dave Chinner wrote: > On Sun, Feb 17, 2013 at 09:20:46PM +0530, Subranshu Patel wrote: >> I created 2 filesystem on my system (RHEL 6.3, kernel 2.6.32) - XFS >> and EXT4 and mounted them. >> >> On both the filesystem I executed a program which performed the following: >> - Create large number of directories >> - Removing all the directories >> >> During the execution of the program, I monitored the %user and %system >> CPU usage (using sar command) >> >> For EXT4 the %user was 1% and %system was 21% >> For XFS the %user was 20% and %system was 52% >> >> But I am not able to understand %user consumption, since %user reports >> the CPU usage of the user level (application), shouldn't it be the >> same for both EXT4 and XFS? > > Perf is your friend. > >> What are the factors that decide the %user component keeping in mind >> that only the filesystem is different? > > Page faults, CPU cache misses. XFS has a significantly large icache > and dcache footprint than ext4, so those are the likely causes. > > Cheers, > > Dave. > -- > Dave Chinner > david@fromorbit.com _______________________________________________ xfs mailing list xfs@oss.sgi.com http://oss.sgi.com/mailman/listinfo/xfs