[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [dm-devel] dm-crypt parallelization patches



> The patches mainly help with sequential I/O (sequential I/O is normally 
> submitted from a few CPUs, so existing dm-crypt doesn't parallelize 
> right). There is a decrease of performance when 12 threads are doing 
> writes on ramdisk:

That's a bit worrying. Could you please find a larger system (4S) in RH and
try with that?

I suspect you may need to limit the load balancing to smaller groups
of CPUs. Querying all other CPUs all the time is unlikely to scale.

-Andi


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]