On Mon, 29 Aug 2005 10:28:08 -0400
"Lewis Adam-CAL022" <Adam.Lewis@xxxxxxxxxxxx> wrote:
> Hi all,
>
> I am evaluating the Westwood protocol in the lab using NISTnet as an
> emulator. So far I am seeing little to zero improvement of TCP
> performance using Westwood as opposed to the Vegas implementation in the
> kernel. I have two NISTnet boxes to simulate packet loss and delay in
> both the upstream and downstream path. NISTnet is programmed to induce
> 500ms delay, drop 5% of packets and emulate 1Mbps bandwidth. I'm using
> iperf and sending 3.2M of data. I am using the 2.6.12 kernel. These
> results are very disappointing as publications I've seen show much
> better improvement with a high BDP (I assume bandwidth=1Mbps *
> delay=500ms qualifies). I am seeing best case 5% improvement over
> Reno/Vegas.
>
> Am I using a valid version of Westwood in this kernel? Is there anyway
> to verify that it is running, aside from cat-ing the /proc files?
> Anything in /var/log/messages or elsewhere? Or is this really the best
> one can hope for with Westwood (e.g. 5% improvement)?
>
Have you reconfigured to support bigger TCP memory usage?
With the default kernel values you will end up being window limited
under high BDP.
Something like this in /etc/sysctl.conf (most distros use this).
# increase Linux TCP buffer limits
net.core.rmem_max = 8388608
net.core.wmem_max = 8388608
# increase Linux autotuning TCP buffer limits
net.ipv4.tcp_rmem = 4096 87380 8388608
net.ipv4.tcp_wmem = 4096 65536 8388608
-
: send the line "unsubscribe linux-net" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
