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 65A227F3F for ; Wed, 16 Oct 2013 10:05:04 -0500 (CDT) Received: from cuda.sgi.com (cuda3.sgi.com [192.48.176.15]) by relay1.corp.sgi.com (Postfix) with ESMTP id 58B908F804B for ; Wed, 16 Oct 2013 08:05:01 -0700 (PDT) Received: from sandeen.net (sandeen.net [63.231.237.45]) by cuda.sgi.com with ESMTP id U2o27lYjNZcVv2bW for ; Wed, 16 Oct 2013 08:05:00 -0700 (PDT) Message-ID: <525EAB1C.6020504@sandeen.net> Date: Wed, 16 Oct 2013 10:05:00 -0500 From: Eric Sandeen MIME-Version: 1.0 Subject: Re: XFS - configuration for multi-thread high-speed streams References: In-Reply-To: List-Id: XFS Filesystem from SGI List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Content-Type: text/plain; charset="windows-1252" Content-Transfer-Encoding: quoted-printable Errors-To: xfs-bounces@oss.sgi.com Sender: xfs-bounces@oss.sgi.com To: Mike Jensen Cc: "xfs@oss.sgi.com" On 10/15/13 3:45 PM, Mike Jensen wrote: > Hi Group - 1^st time post, looks like just use the email address I have s= een from other posts > = > If the setup/info/questions below are not suitable for this list, is ther= e another xfs list available for these?? > = User questions are appropriate here, but since this is a Red Hat distro ker= nel, not upstream, it's probably better to engage Red Hat for information o= n optimal tuning of something like this. -Eric > = > Thanks > = > mike > = > = > = > Configuration =96 > = > = > = > - RH 6.4 on HP G8 server w/16GB mem (could add a lot more) > = > - 12Gb SAS connections (2) to storage > = > - Storage > = > - 96 LFF HDD=92s (4TB 7K NLSAS) presently organized as 8 x 10+2 = R6 raidsets > = > - 8 volumes mapped to host ports (MPIO engaged) > = > - Mdadm used to stripe (/dev/dm*) LUNs into a single md device (= approx. 300+TB file system space) > = > - Mkfs.xfs used to lay down file system > = > - = > = > - Workload > = > - Application creates file of specified length in dir of mount p= oint, IO=92s then issued to file(s), each file receives 1 stream of IO > = > - 1 x high speed stream 1000MB/s Seq W (IO=92s arrive as 512KB -= =3D> 4MB frame would be 8 x 512KB IO=92s) > = > - 2 x med-speed streams 200MB/s Seq W =93 > = > - 30-50 low speed streams 10MB/s Seq W =93 > = > = > = > Objectives/Questions =96 > = > = > = > - Would like to optimize xfs/mount parameters to make maximal us= e of storage assets > = > - Thinking of using 15K HDD R1 set for log files =96 would this = get all/most of metadata or just a subset? Right now seeing metadata write= s arriving with data writes and want to peel them off the 7K HDD=92s > = > - Would using sub-directories for each file ensure that each sub= dir/file would be in it=92s own alloc group? And that would keep the stream= ing data in a separate AG =96 could help the storage more efficiently desta= ge the writes from cache to disk > = > = > = > = > = > _______________________________________________ > xfs mailing list > xfs@oss.sgi.com > http://oss.sgi.com/mailman/listinfo/xfs > = _______________________________________________ xfs mailing list xfs@oss.sgi.com http://oss.sgi.com/mailman/listinfo/xfs