[Gluster-users] how to optimize performance on server side

Udo Waechter udo.waechter at uni-osnabrueck.de
Sun Dec 26 17:10:44 UTC 2010


On 22.12.2010, at 19:21, Gotwalt, P. wrote:

> I have 4 nodes with each 4 disks. I want to have the best performance so
> I will make a stripe of these 4 servers. But how can I get the best
> performance out of each server? 
> 1 - Make a software stripe (with md raid0) over all the internal disks
> and export this as 1 volume, and then
> volume create performance-volume stripe 4 node1:/bigdisk node2:/bigdisk
> node3:/bigdisk node4:/bigdisk
Raid0? Seriously? - You know that 0 Data can be rescued if one of those disks crashes?

We have 12disk RAID6 Storage servers with software RAID. 3 Gbit ports and performance is great.

Adding each single disk via gluster is mainly a lot of typing work. Thus we decided to do the Software raid layer and use its management capabilities, then add glusterfs atop of that.

I guess it depends on you workloads. Depending on the network links between you storage nodes and you client nodes, you might not need to tune a lot for speed. (If you have one single Gbit link and your lan is Gbit then Gbit is your bottleneck and not the raid-pattern....
Here again raid0 would give you no performance gain at all.


:: udo waechter - root at zoide.net :: N 52º16'30.5" E 8º3'10.1"
:: genuine input for your ears: http://auriculabovinari.de 
::                          your eyes: http://ezag.zoide.net
::                          your brain: http://zoide.net

More information about the Gluster-users mailing list