[Gluster-devel] Scaled down a bit.
krishna at zresearch.com
Wed Nov 28 18:15:07 UTC 2007
On Nov 28, 2007 11:11 PM, Chris Johnson <johnson at nmr.mgh.harvard.edu> wrote:
> On Wed, 28 Nov 2007, Kevan Benson wrote:
> > Chris Johnson wrote:
> >> Hi again.
> >> I've scaled everything way back as was suggested. I'm also using
> >> a dd file read as a test rather than iozone. I have one brick on the
> >> server mounted on the client, that's it. Using time dd I get between
> >> 260 to 285 MS to read a 24 MB file. Doing the same thing with NFS I
> >> get about 20 to 30 MS less but I can live with that.
OK, noted. I am guessing this is because of user space overhead.
> >> I tried with and without io-threads on the server end. No
> >> noticable change either way. I've read with io-threads is supposed to
> >> do. I had 8 threads defined. What does it do and does it make a
> >> different which side it's on, client or server?
> > io-threads is supposed to speed up serving of files when multiple files are
> > requested at the same time (so dd won't trigger any benefits from it at all,
> > being a single request). I.e. In situations on the server where it would be
> > blocking on new requests until finished with the current request, io-threads
> > allows another thread to handle the request for better response and
> > performance (at least that's how I understand it).
Correct. When multiple processes request for multiple files, performance gain
will be significant.
> > As another email a few days ago stated, io-threads on the client side will
> > have little or no effect.
I had observed that client side io-threads also improves response time.
> Ah ha! Ok, tnx.
> >> I also tried the io-cache on the client side. MAN does that
> >> work. I had a 256 MB cache defind. A reread of my 24 MB file took 72
> >> MS. I don't think it even bothered with the server much. I need to
> >> try that on the server. Might help if a bunch of computer nodes
> >> hammer on the same file at the same time.
> > Careful with io-cache and io-threads together, depending on where you define
> > it (I think), the cache is per-thread. so if you have 8 threads and a 256 MB
> > cache defined, be prepared for 2 GB of cache use...
No, If you define one io-cache translator there is only one cache. All the
threads will refer to the same io-cache translator with which it is associated
More information about the Gluster-devel