Steffen, What config you use depends a lot of what kind of data usage you have. You need to think if you really need 3 copies of the data. Please use a recent version ( 3.0.3 ). Regards, Tejas. ----- Original Message ----- From: "Steffen Grunewald" <steffen.grunewald@xxxxxxxxxx> To: "GlusterFS mailing list" <gluster-devel@xxxxxxxxxx> Sent: Monday, March 15, 2010 8:20:06 PM GMT +05:30 Chennai, Kolkata, Mumbai, New Delhi Subject: Large cluster data pool from 150 non-redundant disks? Hi, After having had a look into GlusterFS quite a while ago, I'm now planning to use 2.0.9 to implement a large data pool on our cluster. I've got 150 nodes, each with a spare disk. With GlusterFS' self-healing features, I think a triple redundancy would be more than enough, giving me a "RAID-0" over 50 "RAID-1"s consisting of 3 disks (bricks) each. Since a n-fold "RAID-1" (replicate) would mean that a machine writing to the FS would have to split its bandwidth by n, and the total disk capacity is divided by n as well, 3 appears to be a good compromise. Read access would be spread over all disks evenly, and I'm not worried about that right now (it can only get better, compared to 100+ nodes trying to access files from a single server). Is there a catch somewhere that I don't see? Any suggestions which translators to use (besides posix-locks on the server side)? What's the difference between "afr" and "replicate"? Regards, Steffen -- Steffen Grunewald * MPI Grav.Phys.(AEI) * Am M�hlenberg 1, D-14476 Potsdam Cluster Admin * http://pandora.aei.mpg.de/merlin/ * http://www.aei.mpg.de/ * e-mail: steffen.grunewald(*)aei.mpg.de * +49-331-567-{fon:7233,fax:7298} No Word/PPT mails - http://www.gnu.org/philosophy/no-word-attachments.html _______________________________________________ Gluster-devel mailing list Gluster-devel@xxxxxxxxxx http://lists.nongnu.org/mailman/listinfo/gluster-devel
