mosburn wrote:
> I am working on a large install of Zenoss and am trying to figure out the 
> ratio of hardware resources/number of devices. My current server is a beefy 
> dual core 2.4 Ghz, 8gb ram and more then enough disk space that is constantly 
> pegged with 1400 devices being monitored. To relieve some of the load we have 
> set up remote collectors at all of our sites and this helped some, but I 
> regularly see loads of 9 and 10 on the main server. Has any one figured out 
> the correct requirements for how beefy a server needs to be before I turn on 
> syslog?
>   
As with any performance related issue for any application, you must 
first determine *what* is causing the load.  Just throwing hardware or 
new collectors at it may be more trouble than it is worth.

If you have determined that your disks are the bottleneck, you next need 
to determine what process is hogging the disks.  You can use something 
like iotop for this. http://guichaz.free.fr/iotop/

Once you've determined what process is hogging all the disk, you can 
than look at the type of data it is reading/writing.  Is the data 
valuable?  If not, put it on tmpfs.  Is it hogging the disk based on 
throughput, seek times or latency?  Then you can optimize a storage 
solution based upon the characteristics of your data.

Your solution might be as simple as adjusting some caching parameters, 
but you won't know until you know why exactly the disks are being slammed.

Nathaniel
_______________________________________________
zenoss-users mailing list
[email protected]
http://lists.zenoss.org/mailman/listinfo/zenoss-users

Reply via email to