Hi there
I'm doing some testing on raid10 layouts and chunksizes.
I am checking chunksizes 128, 1024 and 8192, on a 6-disk array.
On streaming read / streaming write (dd bs=1M from/to /dev/zero
/dev/null) the speed of 128 and 1024 on my hardware is around 270MB/sec
writes (all layouts) and 280MB/sec(near)--450MB/sec(far) read speed.
But when I try chunksize 8192 speed drops significantly and I have
231(near)--175(far) write speed and 205(near)--323(far) read speed.
I tried the kitchen sink on settings... but there's no way to bring 8192
speed up to the others. I tried raising nr_requests on all the drives up
by a factor 16, changing schedulers, raising deadline scheduler
expirations, readahead on drives... no way.
This surprises me... what could be the reason? In particular on a near
layout I can't see what could be wrong with a big chunksize.
I suspect the situation would have gone worse if I went further up with
the chunksize.
There is plenty of free RAM (8GB free).
Thanks for your insights
--
To unsubscribe from this list: send the line "unsubscribe linux-raid" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
