On Tue, Apr 14, 2009 at 6:39 PM, Robert Hancock <hancockrwd@xxxxxxxxx> wrote:
> James Bottomley wrote:
>>
>> On a final note about the urgency of getting libata out of SCSI: Intel
>> has been worrying for a while about the fatness of the SCSI/libata
>> stack, and its effects on performance, especially command transmission
>> via SAT, so I'm hoping they'll be supporting the effort.
>
> I really don't see this as being a big driver for this move. If you look at
> the code that does the translation of SCSI commands to ATA commands, there
> really is not much there at all of any consequence to CPU usage.
It's measurable. Just need to get sufficient IO rates going. Normal JBOD/RAID
controllers don't care.
See Matthew "the Intel guy" :) Wilcox and Kristen Accardi's paper at LSF2008:
http://iou.parisc-linux.org/lsf2008/IO-latency-Kristen-Carlson-Accardi.pdf
or just google for "Matthew Wilcox SSD perf" - first four hits I
looked at are what I expected:
hardware.slashdot.org/article.pl?sid=09/02/13/2337258&from=rss
markmail.org/message/w22r6y4gik7bjf2w
https://kerneltrap.org/mailarchive/linux-scsi/2008/12/10/4388804/thread
www.usenix.org/events/lsf08/tech/IO_Carlson_Accardi_SATA.pdf
(last one is the official location of the same paper)
> Compared to
> any kind of hardware/controller interactions I wouldn't say it's likely to
> be a significant bottleneck at all. In oprofile runs I've done with heavy
> ATA activity, the top time consumers are the interrupt handlers,
interrupts just introduce completion reporting latency. Interrupt mitigation
techinques/smarter controller can reduce this.
> command issue paths,
Hrm? Is this in the device driver?
> code that actually is poking IO registers.
Stupid controller design. NICs have been able to run without
MMIO *Reads* in the performance for more than 5 years now.
New "Enterprise" SAS/SATA controllers are better but I'm not
at liberty to discuss those. (sorry)
> The libata-scsi code
> hasn't even shown up on the radar in my experience.
It won't for normal disk IOPS rates (<1000 IOPS per disk).
Run it at 20K or 50K IOPS and see again.
NICs are pushing alot more than that.
hth,
grant
--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
