Re: big picture UDP/IP performance question re 2.6.18 -> 2.6.32

2 messages, 2 authors, 2011-10-04 · open the first message on its own page

Re: big picture UDP/IP performance question re 2.6.18 -> 2.6.32

From: Serge Belyshev <hidden>
Date: 2011-10-04 20:12:12

BFS FAQ writes:
NUMA aware?

It is NOT NUMA aware in the sense that it does any fancy shit on NUMA, but
it will work on NUMA hardware just fine. Only the really big NUMA hardware
is likely to suffer in performance, and this is theoretically only, since
no one has that sort of hardware to prove it to me, but it seems almost
certain. 
This part of FAQ entry was written before 0.300, (i.e. more than 2 years ago)

v0.300 onwards have NUMA enhancements.
And this was added after. =)

As of now, BFS scheduling is NUMA-aware. (see resched_best_mask())

Re: big picture UDP/IP performance question re 2.6.18 -> 2.6.32

From: Con Kolivas <hidden>
Date: 2011-10-04 22:32:19

On Wed, 5 Oct 2011 07:12:06 Serge Belyshev wrote:
BFS FAQ writes:
quoted
NUMA aware?

It is NOT NUMA aware in the sense that it does any fancy shit on NUMA,
but it will work on NUMA hardware just fine. Only the really big NUMA
hardware is likely to suffer in performance, and this is theoretically
only, since no one has that sort of hardware to prove it to me, but it
seems almost certain.
This part of FAQ entry was written before 0.300, (i.e. more than 2 years
ago)
quoted
v0.300 onwards have NUMA enhancements.
And this was added after. =)

As of now, BFS scheduling is NUMA-aware. (see resched_best_mask())
Admittedly the wording of this got confused as BFS developed. It originally 
had no NUMA enhancements, but now it does. The code to benefit NUMA is quite 
simple though, but then, so is everything in BFS.

Regards,
Con
-- 
-ck
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help