On Mon, 2012-09-10 at 12:02 +0100, Gordan Bobic wrote:
> On 09/10/2012 01:42 AM, John A. Sullivan III wrote:
>
> >>> If so, I think it would be even better than OpenSolaris
> >>> as it appears the network stack latency is lower in Linux than
> >>> OpenSolaris from what we've seen. Thanks - John
> >>
> >> ZFS doesn't go anywhere near the network stack so I don't know what
> >> exactly the connection here might be.
> >
> > Network latency for IOPS in a SAN. For small block sizes, I would think
> > latency would be even more of an issue than bandwidth when talking to a
> > ZFS based SAN. Then again, I suppose command tag queuing would coalesce
> > those so latency would not be as critical.
>
> Latency (ping time) on a typical Gbit network is < 100us. Latency of a
> 15,000rpm disk is 4ms. That's a 40x difference. I don't see the network
> latency being particularly relevant at between 1-2.5% of your disk
> latency (depending on the speed of your disks).
>
> Gordan
It does when the SAN is serving from cache and the OS is adding almost a
full ms of latency as we found in some of our work with early Nexenta
units. Thanks - John
Received on Mon Sep 10 13:32:18 2012