Received: with ECARTIS (v1.0.0; list xfs); Thu, 08 Jun 2006 07:48:59 -0700 (PDT)
Received: from ty.sabi.co.UK (82-69-39-138.dsl.in-addr.zen.co.uk [82.69.39.138])
	by oss.sgi.com (8.12.10/8.12.10/SuSE Linux 0.7) with ESMTP id k58EmleZ008779
	for <linux-xfs@OSS.SGI.com>; Thu, 8 Jun 2006 07:48:50 -0700
Received: from from [127.0.0.1] (helo=base.ty.sabi.co.UK)
	by ty.sabi.co.UK with esmtp(Exim 4.62 #1)
	id 1FoLIW-0004h7-EX
	for <linux-xfs@OSS.SGI.com>; Thu, 08 Jun 2006 15:15:44 +0100
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <17544.12555.858981.847208@base.ty.sabi.co.UK>
Date: Thu, 8 Jun 2006 15:15:39 +0100
X-Face: SMJE]JPYVBO-9UR%/8d'mG.F!@.,l@c[f'[%S8'BZIcbQc3/">GrXDwb#;fTRGNmHr^JFb
 SAptvwWc,0+z+~p~"Gdr4H$(|N(yF(wwCM2bW0~U?HPEE^fkPGx^u[*[yV.gyB!hDOli}EF[\cW*S
 H<GG"+i\3#fp@@EeWZWBv;]LA5n1pS2VO%Vv[yH?kY'lEWr30WfIU?%>&spRGFL}{`bj1TaD^l/"[
 msn( /TH#THs{Hpj>)]f><W}Ck9%2?H"AEM)+9<@D;3Kv=^?4$1/+#Qs:-aSsBTUS]iJ$6
Precedence: air-mail
To: Linux XFS <linux-xfs@oss.sgi.com>
Subject: Re: separate log and structure from user data device?
In-Reply-To: <20060608104242.I710447@wobbly.melbourne.sgi.com>
References: <Pine.LNX.4.58.0606051402410.18047@kurp.hut.fi>
	<8630.1149517148@ocs3.ocs.com.au>
	<20060606101258.B644608@wobbly.melbourne.sgi.com>
	<Pine.LNX.4.58.0606061553320.31122@kurp.hut.fi>
	<20060608104242.I710447@wobbly.melbourne.sgi.com>
X-Mailer: VM 7.17 under 21.4 (patch 19) XEmacs Lucid
From: pg_xfs@xfs.for.sabi.co.UK (Peter Grandi)
X-Disclaimer: This message contains only personal opinions
X-archive-position: 7908
X-ecartis-version: Ecartis v1.0.0
Sender: xfs-bounce@oss.sgi.com
Errors-to: xfs-bounce@oss.sgi.com
X-original-sender: pg_xfs@xfs.for.sabi.co.UK
Precedence: bulk
X-list: xfs
Content-Length: 684
Lines: 23


[ ... curious complaints about write performance ... ]

>> /* O_DIRECT for "realtime" */

nathans> You don't need O_DIRECT for realtime these days.

>> assert( (fid = open(pn,
>>   O_WRONLY|O_CREAT|O_DIRECT|O_SYNC|O_LARGEFILE)) != -1 );

nathans> Thats a shocking use of assert. :)

Thats too kind :-).

However, I am puzzled as you comment on 'O_DIRECT' but not on
'O_SYNC'. IIRC XFS performance, and in particular the strategy
to used to allocate large extents, is based on hugely delayed
flushing, and 'O_SYNC' would defeat that, resulting perhaps in
the allocation of many small extents, and this might impact
write speed too (more metadata writes). Is this plausible?

[ ... ]