<html>
<head>
<meta content="text/html; charset=ISO-8859-1"
http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<div class="moz-cite-prefix">On 2014-07-22, 3:27 PM, Gilberto Nunes
wrote:<br>
</div>
<blockquote
cite="mid:CAOKSTBscvrYcaxqBB-Dc5vmRNQyG3n7Q59Q2iNhPK+J9XFWwyQ@mail.gmail.com"
type="cite">
<div dir="ltr">
<div>
<div>Forgive me by the off topic, but I wonder if someone here
can point me the performance difference between GlusterFS
and Ceph...</div>
<div>Which one is faster? Why? Some docs or websites??</div>
<div>
Thank you...</div>
<div><br>
</div>
-- <br>
Gilberto Ferreira<br>
</div>
</div>
<br>
<fieldset class="mimeAttachmentHeader"></fieldset>
</blockquote>
I tried Gluster, both native and it's NFS implementation. The NFS
worked great, until it didn't. It began to fail under load. The
Gluster native worked fine, but halved my IO. I was doing replicate
gluster, which sends data to all (in my case 2) bricks over the same
link, essentially reducing your speed. 2 bricks in replicate, half
the speed. 3 bricks, 1/3 of the speed.<br>
<br>
Gluster NFS replicates it's data between servers, not between client
and server, and can be off-loaded to another network link.<br>
<br>
I settled on standard kernel NFS with DRBD to replicate the data. I
changed the Proxmox config files to use NFSv4, and I am super happy.<br>
<br>
Gerald<br>
<br>
</body>
</html>