On Tue, Mar 07, 2006 at 03:03:13PM -0800, David S. Miller ([EMAIL PROTECTED])
wrote:
> From: Evgeniy Polyakov <[EMAIL PROTECTED]>
> Date: Wed, 8 Mar 2006 00:51:33 +0300
>
> > RCU introduces noticeble latencies for writing part compared to
> > rwlock, so workloads when a lot of new short connections are created
> > and destroyed, like busy web server within LAN with small latencies,
> > will suffer from this, doesn't it?
>
> To a certain extent, yes, and that is why this kind of patch
> needs a lot of testing.
>
> But to be honest, TCP is a lot less sensitive to the RCU
> latencies than things like the routing cache are, so I really
> don't expect any real problems in this area.
>
> You cannot create one full-fledged TCP socket per packet,
> but you can create a new routing cache entry per packet :-)
When I tested RCU for similar change for kevent, but postponed more work
to RCU callback, including socket closing and some attempts to inode
dereferencing, such change forced performance degradation for httperf
benchmark and trivial web server from 2600 requests/sec down to 2200
with 4.1K index page and 1500 MTU.
I still have this testbed and will gladly test this change tomorrow
if it applies cleanly to 2.6.15-rc7?
--
Evgeniy Polyakov
-
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
