[DRBD-user] read-balancing: could not measure any performance benefits + alternative proposal

Arnold Krille arnold at arnoldarts.de
Fri Jan 6 14:47:20 CET 2012

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


On Friday 06 January 2012 09:10:02 Felix Frank wrote:
> On 01/05/2012 06:22 PM, Lutz Vieweg wrote:
> > But I hoped for some increased performance when running two
> > independent processes reading two different files.
> 
> in most load balancing scenarios, I wouldn't expect to notice any
> performance differences between one and two clients respectively.
> 
> I would include as many clients in any load test as the system under
> test can sensibly service. In the case of disk performance, this will
> not be easy, seeing as you need to be careful not to measure your page
> cache performance instead of disk performance etc.
> 
> Still, I believe you may get better results if you can find some "true
> world" application that is both multi threaded and I/O bound, and have
> it do some heavy lifting with and without disk load balancing.
> 
> MySQL comes to mind, if you drastically limit its in-memory caches.

dbench would be my current weapon of choice. And then do a test of 2^x (with 
x=[1,10]) clients.

Arnold
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20120106/03dda0c9/attachment.pgp>


More information about the drbd-user mailing list