<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
On 06/06/2011 01:34 AM, Jonathan Collingridge wrote:
<blockquote cite="mid:4DEC1276.4020609@msp.com.au" type="cite">
<meta http-equiv="content-type" content="text/html;
charset=ISO-8859-1">
Hi,<br>
Ive read just about everything i can find out there on Gluster,
and waded through a lot of the old posts on the user-list. I'm
wondering what the current state of gluster is for interop. with
windows 7 clients. <br>
<br>
I work for a company that deals with images - a lot of them are
jpegs around 2Mb and raw camera images at 14-20Mb. We generate
about 4-6 Tb of these each year.<br>
<br>
Currently we use expensive (fibre channel disk) block based
storage over 4G fibre SAN; attached to this is a Ployserve (a HP
product now) active/active cluster of 4 windows servers. All of
this is due for a refresh this year (Servers with faster I/O and
10GE network) and i'm looking at all the alternatives, HP have an
(IBRIX sourced) array with 2 front end servers, runs redhat,
feature set similar to gluster....<br>
<br>
I'm wondering you have any experience with a Gluster config, with
a similar workload? What about windows 7 enterprise clients - use
NFS or Samba?<br>
<br>
I'm considering gluster nodes as HP DL180G6 with 25x 2.5" 300Gb
10K SAS Drives in each (to keep the number of spindles up). A pair
of 10GbE cards for connectivity. Do you have any thoughts on this
possible config? If you were building a Gluster system from
scratch with an Medium/Large business budget, how would you do it?<br>
<br>
Thanks for any help you can give with my questions.<br>
</blockquote>
<br>
hi!<br>
<br>
We use glusterfs in a similar environment (VFX) and it works:<br>
<br>
- 5 nodes, each has 10x750 7200 RPM SATA HDDs<br>
- GE network (with bonding in the cluster nodes)<br>
- samba<br>
- at this time we are running Gluster Storage Platform (FC11) and we
are planning to move to glusterfs 3.1<br>
<br>
It works with samba pretty well, but could not achieve the maximum
speed with one instance. We get the best speed when we use the nodes
parallell way. So on the (render) clients we map shares from all the
nodes and multiple render process is working on machines at the same
time. This way the speed grows up almost linearly.<br>
<br>
tamas<br>
</body>
</html>