Gluster users - Jul 2015 - GlusterFS getattr performance Apache static

Hi,
I'm trying to setup a GlusterFS cluster on AWS for hosting static content
which will be served by an Apache server.
Gluster will host about a very large amount of small files summing up to
couple of TB.

In this scenario almost 90% of each IO is of type "getattr" in charge
of
checking static file date ( apache last modified ) and I'm expecting a lot
of IOPS ( getattr).

I've already planned to run Gluster server on several nodes to split load
and on performant disks to handle this load. Choose filesystem is XFS .

Here are my questions:
 1. How does GlusterFS handles getattr requests? reading from disks or
served by server cache?
 2. Do you suggest to use NFS or Gluster on Client side ?
 3. To ensure cluster will sustain the load should I increase disk IO or
Server resources CPU RAM?
 4. In case of getattr request on server 1, does gluster server node check
its content only or does it query all other replica ?


Many thanks for your answers.
GPic
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://www.gluster.org/pipermail/gluster-users/attachments/20150706/1f093ca5/attachment.html>