[OpenAFS] Heavy performance loss on gigabit ethernet

[Nickolai Zeldovich]

On Thu, 2004-08-19 at 00:38 +0200, Enrico Scholz wrote:
> As you can see, some datagrams of jumbo-packet #58307 get lost (probably
> an expected 1000Mbs -> 100Mbs effect). Then, after packet #58312 a see a
> 0.35 seconds delay and the transport continues.

Rx wasn't really designed for fast networks :-)  Try this patch.

--- src/rx/rx.c 18 Aug 2004 22:46:37 -0000      1.59
+++ src/rx/rx.c 18 Aug 2004 22:54:17 -0000
@@ -5603,7 +5603,7 @@
      * be switched and/or swapped out.  So on fast, reliable networks, the
      * timeout would otherwise be too short.
      */
-    rtt_timeout = (peer->rtt >> 3) + peer->rtt_dev + 350;
+    rtt_timeout = (peer->rtt >> 3) + peer->rtt_dev;
     clock_Zero(&(peer->timeout));
     clock_Addmsec(&(peer->timeout), rtt_timeout);


-- kolya