* RPS runs with multiqueue NIC
@ 2011-02-22 3:07 Jon Zhou
2011-02-22 5:46 ` Eric Dumazet
0 siblings, 1 reply; 2+ messages in thread
From: Jon Zhou @ 2011-02-22 3:07 UTC (permalink / raw
To: netdev@vger.kernel.org
Hi
What will happen if RPS runs on kernel 2.6.36.4 machine where a multiqueue NIC intel x520 installed
Will it decrease or improve performance?
rui
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: RPS runs with multiqueue NIC
2011-02-22 3:07 RPS runs with multiqueue NIC Jon Zhou
@ 2011-02-22 5:46 ` Eric Dumazet
0 siblings, 0 replies; 2+ messages in thread
From: Eric Dumazet @ 2011-02-22 5:46 UTC (permalink / raw
To: Jon Zhou; +Cc: netdev@vger.kernel.org
Le lundi 21 février 2011 à 19:07 -0800, Jon Zhou a écrit :
> Hi
>
> What will happen if RPS runs on kernel 2.6.36.4 machine where a multiqueue NIC intel x520 installed
> Will it decrease or improve performance?
It depends on the workload.
For typical TCP flows, hardware multiqueue is generally better,
especially if the NIC supports the ethtool -X command (so that you can
tune effective number of cpus dedicated to the softirq processing).
Packet are delivered directly to the CPU target, without taking an extra
IPI.
For UDP workloads, adding RPS might be a win, because many NIC dont take
into account ports numbers in their hash computation, while
net/core/dev.c hash function does. If you receive trafic coming from a
single remote machine, RPS helps to spread flows between several cpus.
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2011-02-22 5:54 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2011-02-22 3:07 RPS runs with multiqueue NIC Jon Zhou
2011-02-22 5:46 ` Eric Dumazet
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.