On 6/1/2011 2:21 PM, Alan Cox wrote:
Thanks, Asai! I don't think cfq is the ideal I/O scheduler to be
testing. Could you run again with deadline and/or noop and see how that
changes your throughput and perf report? Also, just for completeness,
could you tell us which kernel you ran this against?
kernel 2.6.38.6
How many processors is this system, just looking at the lock contention
which is pretty horrible.
8 processors (2 quad-core CPUs)
Intel(R) Xeon(R) CPU X5672 @ 3.20GHz
I'd been expecting various red flags in the AHCI/libata/scsi queue code
but it seems at first glance that the block queue stuff is killing us and
the scsi/ata code is a distraction (unless of course its causing a lot of
the lock time)
No-op would be most interesting but the I/O scheduler numbers don't look
pretty.
Alan
On looking into the data in below links, lock and block queue are
consuming more time when running with ahci driver. Correct me if I
missing something.
With mtipx2xx, the driver is spot on processing I/O. Any idea of who is
the top offender when running with mtipx2xx?
Filename | Description
-------------------------------------------------------------------------------------------
perf_report_ahci_cfq : perf call graph for ahci driver with cfq
scheduler enabled
http://www.micron.com/get-document/?documentId=6768
vdbench.ahci.cfq.html : vdbench summary of a run for ahci driver
with cfq scheduler enabled
http://www.micron.com/get-document/?documentId=6774
perf_report_ahci_deadline : perf call graph for ahci driver with
deadline scheduler enabled
http://www.micron.com/get-document/?documentId=6769
vdbench_ahci.deadline.html: vdbench summary of a run for ahci driver
with deadline scheduler enabled
http://www.micron.com/get-document/?documentId=6775
perf_report_ahci_noop : perf call graph for ahci driver with noop
scheduler enabled
http://www.micron.com/get-document/?documentId=6770
vdbench_ahci.noop.html : vdbench summary of a run for ahci driver
with noop scheduler enabled
http://www.micron.com/get-document/?documentId=6776
perf_report_micron : perf call graph for Micron block driver
(mtipx2xx)
http://www.micron.com/get-document/?documentId=6772
vdbench_micron.html : vdbench summary of a run for Micron block
driver (mtipx2xx)
http://www.micron.com/get-document/?documentId=6777
This is the first time we are hosting files for public access, that
process took time to get to here.
--
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
