Re: Slow disks.

[Jeff Moyer <jmoyer@redhat.com>]

Rogier Wolff <R.E.Wolff@BitWizard.nl> writes:

> On Wed, Dec 22, 2010 at 11:27:20AM -0500, Jeff Moyer wrote:
>> Rogier Wolff <R.E.Wolff@BitWizard.nl> writes:
>> 
>> > Unquoted text below is from either me or from my friend. 
>> >
>> >
>> > Someone suggested we try an older kernel as if kernel 2.6.32 would not
>> > have this problem. We do NOT think it suddenly started with a certain
>> > kernel version. I was just hoping to have you kernel-guys help with
>> > prodding the kernel into revealing which component was screwing things
>> > up....
>> [...]
>> > ata3.00: ATA-8: WDC WD10EARS-00Y5B1, 80.00A80, max UDMA/133
>> 
>> This is an "Advanced format" drive, which, in this case, means it
>> internally has a 4KB sector size and exports a 512byte logical sector
>> size.  If your partitions are misaligned, this can cause performance
>> problems.
>
> This would mean that for a misalgned write, the drive would have to
> read-modify-write every super-sector. 
>
> In my performance calculations, 10ms average seek (should be around
> 7), 4ms average rotational latency for a total of 14ms. This would
> degrade for read-modify-write to 10+4+8 = 22ms. Still 10 times better
> than what we observe: service times on the order of 200-300ms. 

I didn't say it would account for all of your degradation, just that it
could affect performance.  I'm sorry if I wasn't clear on that.

>  > md1 : active raid5 sda2[0] sdd2[3](S) sdb2[1] sdc2[4]
>> >       39067648 blocks super 1.2 level 5, 512k chunk, algorithm 2 [3/3]
>> > [UUU]
>> 
>> A 512KB raid5 chunk with 4KB I/Os?  That is a recipe for inefficiency.
>> Again, blktrace data would be helpful.
>
> Where did you get the 4kb IOs from? You mean from the iostat -x
> output?

Yes, since that's all I have to go on at the moment.

> The system/filesystem decided to do those small IOs. With the
> throughput we're getting on the filesystem, it better not try to write
> larger chuncks...

Your logic is a bit flawed, for so many reasons I'm not even going to
try to enumerate them here.  Anyway, I'll continue to sound like a
broken record and ask for blktrace data.

> I have benchmarked my own "high bandwidth" raid arrays. I benchmarked
> them with 128k, 256, 512 and 1024k blocksize. I got the best
> throughput (for my benchmark: dd if=/dev/md0 of=/dev/null bs=1024k)
> with 512k blocksize. (and yes that IS a valid benchmark for my
> usage of the array.)

Sorry, I'm not sure I understand how this is relevant.  I thought we
were troubleshooting a problem on someone else's system.  Further, the
window into the workload we saw via iostat definitely shows that smaller
I/Os are issued.

Anyway, it will be much easier to debate the issue once the blktrace
data is gathered.

Happy holidays.

-Jeff