From mboxrd@z Thu Jan 1 00:00:00 1970 From: Suresh Jayaraman Subject: Re: [PATCH 00/09] cifs: local caching support using FS-Cache Date: Thu, 15 Jul 2010 18:23:19 +0200 Message-ID: <4C3F35F7.8060408@suse.de> References: <1278333663-30464-1-git-send-email-sjayaraman@suse.de> <4C3DF6BF.3070001@gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: QUOTED-PRINTABLE Cc: Scott Lovenberg , linux-cifs-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-fsdevel-u79uwXL29TY76Z2rM5mHXA@public.gmane.org, linux-cachefs-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org, David Howells To: Steve French Return-path: In-Reply-To: Sender: linux-cifs-owner-u79uwXL29TY76Z2rM5mHXA@public.gmane.org List-ID: On 07/14/2010 08:09 PM, Steve French wrote: > On Wed, Jul 14, 2010 at 12:41 PM, Scott Lovenberg > wrote: >> On 7/5/2010 8:41 AM, Suresh Jayaraman wrote: >>> >>> This patchset is a second try at adding persistent, local caching f= acility >>> for >>> CIFS using the FS-Cache interface. >>> >>> >> >> Just wondering, have you bench marked this at all? =EF=BF=BDI'd be i= nterested to see >> how this compares (performance and scaling) to an oplock-centric des= ign. >> Yes, I have done a few performance benchmarks with the cifs client (and not SMB2) and I'll post them early nextweek when I'm back (as I'm travelling now). However, I have never done scalability tests (not sure whether there is a way to simulate a number of cifs clients). >> I'd hazard a guess that with pipelining support in SMB2 the performa= nce will >> be even better since you can have a hot cache and more requests in f= light. >=20 > Yes - very plausibly >=20 I have not tried the new SMB2 client. But, it seems the pipelining support, Oplocks (only Level II kind) could help improve performance. Thanks, --=20 Suresh Jayaraman