Re: stochastic fair queueing in the elevator [Re: [BENCHMARK] 2.4.20-ck3

[Nick Piggin]

Andrew Morton wrote:

>Hans Reiser <[e-mail]> wrote:
>
>>readahead seems to be less effective for non-sequential objects.  Or at 
>>least, you don't get the order of magnitude benefit from doing only one 
>>seek, you only get the better elevator scheduling from having more 
>>things in the elevator, which isn't such a big gain.
>>
>>For the spectators of the thread, one of the things most of us learn 
>>from experience about readahead is that there has to be something else 
>>trying to interject seeks into your workload for readahead to make a big 
>>difference, otherwise a modern disk drive (meaning, not 30 or so years 
>>old) is going to optimize it for you anyway using its track cache.
>>
>>
>
>The biggest place where readahead helps is when there are several files being
>read at the same time.  Without readahead the disk will seek from one file
>to the next after having performed a 4k I/O.  With readahead, after performing
>a (say) 128k I/O.  It really can make a 32x difference.
>
>Readahead is brute force.  The anticipatory scheduler solves this in a
>smarter way.
>
>Yes, device readahead helps.  But I believe the caches are only 4-way
>segmented or thereabouts, so it is easy to thrash them.  Let's test:
>
[snip results]

>
>
>Still only 300%.  I'd expect to see much larger differences on some older
>SCSI disks I have here.
>
It would be interesting to see numbers with and without anticipatory
scheduling as well. I have tried two streaming readers.
Speed with default readahead 13MB/s
Speed with no readahead, no anticipation 7MB/s
Speed with no readahead, anticipation 11.8MB/s

I suspect more readers will simply favour the anticipatory scheduler
more.

Nick