From mboxrd@z Thu Jan 1 00:00:00 1970 From: Bart Van Assche Subject: Re: SRPT and SCST Date: Sun, 30 May 2010 10:01:01 +0200 Message-ID: References: <3142CEFB1403044F9954E2DF6C85660FBB34BD@orca.penguincomputing.com> <3142CEFB1403044F9954E2DF6C85660FBB34BF@orca.penguincomputing.com> <654FA770A883FB43BAF3CB0B1E1DAC8C01C8C4DD@orca.penguincomputing.com> <4AF29201.6000606@penguincomputing.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: QUOTED-PRINTABLE Return-path: In-Reply-To: <4AF29201.6000606-pabcTyWEv4ZW60MLeMDbCVaTQe2KTcn/@public.gmane.org> Sender: linux-rdma-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org To: Philip Pokorny Cc: scst-devel-5NWGOfrQmneRv+LV9MX5uipxlwaOVQ5f@public.gmane.org, linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, Arend Dittmer List-Id: linux-rdma@vger.kernel.org (resending as plain text) On Thu, Nov 5, 2009 at 10:51 AM, Philip Pokorny wrote: > > [ ... ] > > Performance (when it's working) is generally good at 800MB/sec aggreg= ate, but we'd like to see better. =A0It appeared we were getting 1.3GB/= s at one point. (replying to an e-mail that was posted several months ago) You might have hit a limitation of Linux' asynchronous I/O subsystem (at the SRP initiator side). Asynchronous I/O performance has been improved significantly in the 2.6.33 kernel. The throughput I measured with fio, QDR HCA's and a 2.6.34 initiator is as follows: * 940 MB/s for asynchronous writes and 1240 MB/s for asynchronous reads and a block size of 64 KB. * 2200 MB/s for direct writes and 2900 MB/s for direct reads and a block size of 64000 KB. Bart. -- To unsubscribe from this list: send the line "unsubscribe linux-rdma" i= n the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org More majordomo info at http://vger.kernel.org/majordomo-info.html