<html><body>
<p><tt>Hi, Eli,</tt><br>
<tt><br>
> Hi,<br>
> I have a draft implementation of NAPI in ipoib and got the following<br>
> results:<br>
> System descriptions<br>
> ===================<br>
> Quad CPU E64T 2.4 Ghz<br>
> 4 GB RAM<br>
> MT25204 Sinai HCA<br>
> I used netperf for benchmarking, the BW test ran for 600 seconds with 8<br>
> clients and 8 servers.<br>
> <br>
> The results I received are bellow:<br>
> <br>
> netperf TCP_STREAM:<br>
> BW [MByte/sec] clients side [irqs/sec] server side [irqs/sec]<br>
> -------------- ----------------------- ----------------------<br>
> without NAPI: 506 86441 66311<br>
> with NAPI: 550 6830 13600 <br>
> <br>
> <br>
> netperf TCP_RR:<br>
> rate [tran/sec]<br>
> ---------------<br>
> without NAPI: 39600<br>
> with NAPI: 39470<br>
> <br>
> <br>
> <br>
> Please note this is still under work and we plan to do more tests and<br>
> measure on other devices.<br>
</tt><br>
<tt>NAPI patch moves ipoib poll from hardware interrupt context to softirq context. It would reduce the hardware interrupts, reduce hardware latency and induce some network latency. It might reduce cpu utilization. But I still question about the BW improvement. I did see various performance with the same test under the same condition.</tt><br>
<br>
<tt>Have you tested this patch with different message sizes, different socket sizes? Are these results consistent better?</tt><br>
<br>
Thanks<br>
Shirley Ma<br>
IBM Linux Technology Center<br>
15300 SW Koll Parkway<br>
Beaverton, OR 97006-6063<br>
Phone(Fax): (503) 578-7638</body></html>