[Gluster-users] Major performance problem with WRF

[Matt Lawrence]

Anand Avati wrote:
>>> WRF is not the only thing running on the cluster.
>>>
>>> Direct I/O of 1GB takes 2.7 seconds.  NFS takes 17 seconds.  Gluster takes
>>> 100 seconds.
>>>
>>>       
>> Pulled those translators out and it reduced the glusterfs time to 77
>> seconds.  A significant improvement, but still far behind NFS.  Any other
>> suggestions?
>>     
>
> Do you know what block size the operations are performed? Is there a
> way to tune (increase) it? Fuse has a high overhead for small block
> sizes (< 4KB). Alternately, you can try to LD_PRELOAD the booster
> module to your application before starting it which bypasses fuse for
> read/write calls.
>   

For these tests, I used a 16k blocksize and wrote 1GB

-- Matt