From: James Bottomley <James.Bottomley@steeleye.com>
To: "Justin T. Gibbs" <gibbs@scsiguy.com>
Cc: James Bottomley <James.Bottomley@SteelEye.com>,
Jens Axboe <axboe@suse.de>, Matthew Jacob <mjacob@feral.com>,
"Pedro M. Rodrigues" <pmanuel@myrealbox.com>,
Mathieu Chouquet-Stringer <mathieu@newview.com>,
linux-scsi@vger.kernel.org, linux-kernel@vger.kernel.org
Subject: Re: Warning - running *really* short on DMA buffers while doing file transfers
Date: Fri, 27 Sep 2002 13:21:29 -0400 [thread overview]
Message-ID: <200209271721.g8RHLTn05231@localhost.localdomain> (raw)
In-Reply-To: Message from "Justin T. Gibbs" <gibbs@scsiguy.com> of "Fri, 27 Sep 2002 10:26:47 MDT." <2441376224.1033144007@aslan.btc.adaptec.com>
> Which part of the OS are you talking about?
I'm not, I'm talking about the pure physical characteristics of the transport
bus.
> I also do not believe that the command overhead is as significant as
> you suggest. I've personally seen a non-packetized SCSI bus perform
> over 15K transactions per-second.
Well, lets assume the simplest setup possible: select + tag msg + 10 byte
command + disconnect + reselect + status; that's 17 bytes async. The maximum
bus speed async narrow is about 4Mb/s, so those 17 bytes take around 4us to
transmit. On a wide Ultra2 bus, the data rate is about 80Mb/s so it takes
50us to transmit 4k or 800us to transmit 64k. However, the major killer in
this model is going to be disconnection delay at around 200us (dwarfing
arbitration delay, bus settle time etc). For 4k packets you spend about 3
times longer arbitrating for the bus than you do transmitting data. For 64k
packets it's 25% of your data transmit time in arbitration. Your theoretical
throughput for 4k packets is thus 20Mb/s. In my book that's a significant
loss on an 80Mb/s bus.
On Fabric busses, you move to the network model and collision probabilities
which increase as the packet size goes down.
gibbs@scsiguy.com said:
> Because of read-ahead, the OS should never send 16 4k contiguous reads
> to the I/O layer for the same application.
read ahead is basically a very simplistic form of I/O scheduling.
> Hooks for sending ordered tags have been in the aic7xxx driver, at
> least in FreeBSD's version, since '97. As soon as the Linux cmd
> blocks have such information it will be trivial to have the aic7xxx
> driver issue the appropriate tag types.
They already do in 2.5, see scsi_populate_tag_msg() in scsi.h. This assumes
you're using the generic tag queueing, which the aic7xxx doesn't, but you
could easily key the tag type off REQ_BARRIER.
> But this misses the point. Andrew's original speculation was that
> writes were "passing reads" once the read was submitted to the drive.
The speculation is based on the observation that for transactions consisting
of multiple writes and small reads, the reads take a long time to complete.
That translates to starving a read in favour of a bunch of contiguous writes.
I'm sure we've all seen SCSI drives indulge in this type of unfair behaviour
(it does make sense to keep servicing writes if they're direct follow on's
from the previously serviced ones).
> I would like to understand the evidence behind that assertion since
> all drive's I've worked with automatically give a higher priority to
> read traffic than writes since writes can be buffered but reads
> cannot.
The evidence is here:
http://marc.theaimsgroup.com/?l=linux-kernel&m=103302456113997&w=1
James
next prev parent reply other threads:[~2002-09-27 17:16 UTC|newest]
Thread overview: 60+ messages / expand[flat|nested] mbox.gz Atom feed top
2002-09-26 3:27 Warning - running *really* short on DMA buffers while doing file transfers Mathieu Chouquet-Stringer
2002-09-26 6:14 ` Jens Axboe
2002-09-26 7:04 ` Pedro M. Rodrigues
2002-09-26 15:31 ` Justin T. Gibbs
2002-09-27 6:13 ` Jens Axboe
2002-09-27 6:33 ` Matthew Jacob
2002-09-27 6:36 ` Jens Axboe
2002-09-27 6:50 ` Matthew Jacob
2002-09-27 6:56 ` Jens Axboe
2002-09-27 7:18 ` Matthew Jacob
2002-09-27 7:24 ` Jens Axboe
2002-09-27 7:29 ` Matthew Jacob
2002-09-27 7:34 ` Matthew Jacob
2002-09-27 7:45 ` Jens Axboe
2002-09-27 8:37 ` Matthew Jacob
2002-09-27 10:25 ` Jens Axboe
2002-09-27 12:18 ` Matthew Jacob
2002-09-27 12:54 ` Jens Axboe
2002-09-27 13:30 ` Justin T. Gibbs
2002-09-27 14:26 ` James Bottomley
2002-09-27 14:33 ` Jens Axboe
2002-09-27 16:26 ` Justin T. Gibbs
2002-09-27 17:21 ` James Bottomley [this message]
2002-09-27 18:56 ` Justin T. Gibbs
2002-09-27 19:07 ` Warning - running *really* short on DMA buffers while doingfile transfers Andrew Morton
2002-09-27 19:16 ` Justin T. Gibbs
2002-09-27 19:36 ` Warning - running *really* short on DMA buffers while doingfiletransfers Andrew Morton
2002-09-27 19:52 ` Justin T. Gibbs
2002-09-27 21:13 ` James Bottomley
2002-09-27 21:18 ` Matthew Jacob
2002-09-27 21:23 ` James Bottomley
2002-09-27 21:29 ` Justin T. Gibbs
2002-09-27 21:32 ` Matthew Jacob
2002-09-27 22:08 ` Mike Anderson
2002-09-30 23:49 ` Doug Ledford
2002-09-27 21:28 ` Justin T. Gibbs
2002-09-28 15:52 ` James Bottomley
2002-09-28 23:25 ` Luben Tuikov
2002-09-29 2:48 ` James Bottomley
2002-09-30 8:34 ` Jens Axboe
2002-09-29 4:00 ` Justin T. Gibbs
2002-09-29 15:45 ` James Bottomley
2002-09-29 16:49 ` [ getting OT ] " Matthew Jacob
2002-09-30 19:06 ` Luben Tuikov
2002-09-30 23:54 ` Doug Ledford
2002-09-27 19:58 ` Andrew Morton
2002-09-27 20:58 ` Warning - running *really* short on DMA buffers while doing file transfers Justin T. Gibbs
2002-09-27 21:38 ` Patrick Mansfield
2002-09-27 22:08 ` Justin T. Gibbs
2002-09-27 22:28 ` Patrick Mansfield
2002-09-27 22:48 ` Justin T. Gibbs
2002-09-27 18:59 ` Warning - running *really* short on DMA buffers while doingfile transfers Andrew Morton
2002-09-27 14:30 ` Warning - running *really* short on DMA buffers while doing file transfers Jens Axboe
2002-09-27 17:19 ` Justin T. Gibbs
2002-09-27 18:29 ` Rik van Riel
2002-09-27 14:56 ` Rik van Riel
2002-09-27 15:34 ` Matthew Jacob
2002-09-27 15:37 ` Jens Axboe
2002-09-27 17:20 ` Justin T. Gibbs
2002-09-27 12:28 ` Pedro M. Rodrigues
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=200209271721.g8RHLTn05231@localhost.localdomain \
--to=james.bottomley@steeleye.com \
--cc=axboe@suse.de \
--cc=gibbs@scsiguy.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-scsi@vger.kernel.org \
--cc=mathieu@newview.com \
--cc=mjacob@feral.com \
--cc=pmanuel@myrealbox.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).