Subject: Re: Higher than expected disk write(2) latency
From: Chris Mason <chris.mason@oracle.com>
To: Arjan van de Ven <arjan@infradead.org>
Cc: Martin Lucina <mato@kotelna.sk>, Andrew Morton <akpm@linux-foundation.org>,
       Martin Sustrik <sustrik@fastmq.com>, linux-kernel@vger.kernel.org,
       linux-aio@kvack.org
In-Reply-To: <20080710070151.5fccbbcf@infradead.org>
References: <20080628121131.GA14181@nodbug.moloch.sk>
	 <20080709222701.8eab4924.akpm@linux-foundation.org>
	 <4875C45C.2010901@fastmq.com>
	 <20080710011417.95532d51.akpm@linux-foundation.org>
	 <1215696541.24425.214.camel@think.oraclecorp.com>
	 <20080710134144.GA31461@dezo.moloch.sk>
	 <20080710070151.5fccbbcf@infradead.org>
Content-Type: text/plain
Date: Thu, 10 Jul 2008 10:18:32 -0400
Message-Id: <1215699512.24425.218.camel@think.oraclecorp.com>
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1389
Lines: 38

On Thu, 2008-07-10 at 07:01 -0700, Arjan van de Ven wrote:
> On Thu, 10 Jul 2008 15:41:46 +0200
> Martin Lucina <mato@kotelna.sk> wrote:
> 
> > chris.mason@oracle.com said:
> > > Is NCQ enabled on the drive?  The basic way to fix this is to have
> > > multiple requests in flight, which isn't going to happen on sata
> > > with the cache off and with ncq off.
> > 
> > ata1: SATA link up 3.0 Gbps (SStatus 123 SControl 300)
> > ata1.00: ATA-7: Hitachi HDS721616PLA380, P22OAB3A, max UDMA/133
> > ata1.00: 321672960 sectors, multi 0: LBA48 NCQ (depth 31/32)
> > ata1.00: configured for UDMA/133
> > 
> > I'm assuming the above indicates NCQ is enabled.
> > 
> > We'll try running the tests tomorrow with different i/o schedulers and
> > report back.
> 
> 
> another thing to try is using AHCI mode instead; AHCI tends to be
> higher performance
> 
> (and.. for fun try to run latencytop during a run, to see if maybe
> there are unsuspected delay causes)
> 
Any latencies during submit (at least for kernel aio) are probably from
get_request_wait, but the numbers so far look like they are only seeing
one request in flight at a time.

-chris


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/