<HTML>
<HEAD>
<TITLE>Re: RFT: virtio_net: limit xmit polling</TITLE>
</HEAD>
<BODY>
<FONT FACE="Calibri, Verdana, Helvetica, Arial"><SPAN STYLE='font-size:11pt'><BR>
<BR>
<BR>
On 6/29/11 1:42 AM, "Michael S. Tsirkin" <<a href="mst@redhat.com">mst@redhat.com</a>> wrote:<BR>
</SPAN></FONT><BLOCKQUOTE><FONT FACE="Calibri, Verdana, Helvetica, Arial"><SPAN STYLE='font-size:11pt'><BR>
>roprabhu, Tom,<BR>
><BR>
>Thanks very much for the testing. So on the first glance<BR>
>one seems to see a significant performance gain in V0 here,<BR>
>and a slightly less significant in V2, with V1<BR>
>being worse than base. But I'm afraid that's not the<BR>
>whole story, and we'll need to work some more to<BR>
>know what really goes on, please see below.<BR>
><BR>
><BR>
>Some comments on the results: I found out that V0 because of mistake<BR>
>on my part was actually almost identical to base.<BR>
>I pushed out virtio-net-limit-xmit-polling/v1a instead that<BR>
>actually does what I intended to check. However,<BR>
>the fact we get such a huge distribution in the results by Tom<BR>
>most likely means that the noise factor is very large.<BR>
><BR>
><BR>
>From my experience one way to get stable results is to<BR>
>divide the throughput by the host CPU utilization<BR>
>(measured by something like mpstat).<BR>
>Sometimes throughput doesn't increase (e.g. guest-host)<BR>
>by CPU utilization does decrease. So it's interesting.<BR>
><BR>
><BR>
>Another issue is that we are trying to improve the latency<BR>
>of a busy queue here. However STREAM/MAERTS tests ignore the latency<BR>
>(more or less) while TCP_RR by default runs a single packet per queue.<BR>
>Without arguing about whether these are practically interesting<BR>
>workloads, these results are thus unlikely to be significantly affected<BR>
>by the optimization in question.<BR>
><BR>
>What we are interested in, thus, is either TCP_RR with a -b flag<BR>
>(configure with --enable-burst) or multiple concurrent<BR>
>TCP_RRs.<BR>
<BR>
ok sounds good. I am testing your v1a patch. Will try to get some results out end of this week. Thanks.<BR>
<BR>
</SPAN></FONT></BLOCKQUOTE>
</BODY>
</HTML>