[Lustre-discuss] Single client performance
David Simas
dgs at slac.stanford.edu
Thu Mar 11 20:55:17 PST 2010
Hello,
I've got a Lustre system where I'd like to improve single client write
performance, The Lustre manual reads "Lustre single client node throughput
has been measured at 2 GB/s (max)" (section 1.1.1 Lustre Key Features),
and the web site
http://docs.hp.com/en/SFSUGG3-D/apas02.html?btnPrev=%AB%A0prev
indicates single client write performance of around 400 MB/s. On my
system, I can barely get 100 MB/s for writes (measured with iozone).
I'm using Lustre 1.8.2-ext4 on RHEL5 x86_64. I've got four OSSs each
with one OST, in hardware RAID 6. Lustre runs on a 10 Gb network
between the servers and clients. For the OSSs, iozone tells me that
I can write into ext4 on the RAID arrays I configured at 750 MB/s.
Indeed, the write peformance of ext4 vs. ext3 is one reason I'm using
1.8.2-ext4. I can also confirm (using ttcp) that the 10 Gb network
between the clients and severs can do 750 MB/s. So I don't think that
the network or disk/controller/RAID configurations are holding me
back.
I've tried stripping across 1, 2, 3, and 4 OSTs, without much difference
in results. I've tried disabling synchronous journals on the OSSs, and
disable the OSTs' write-through cache, again to little effect. Should
I try the client I/O RPC stream tunables next?
The goal here is to use a Lustre file system as a sort of buffer in a
data acquisition system. The file system needs to support 4 - 24 input
streams writing at 250 MB/s. I realize that Lustre wasn't designed for
this sort of workload. But if I could get it to work, it would simplify
the data acquisition pipeline.
I'm grateful for any advice.
David Simas
system.
More information about the lustre-discuss
mailing list