<meta http-equiv="Content-Type" content="text/html; charset=utf-8"><meta name="ProgId" content="Word.Document"><meta name="Generator" content="Microsoft Word 11"><meta name="Originator" content="Microsoft Word 11"><link rel="File-List" href="file:///C:%5CDOCUME%7E1%5Cwp%5CLOCALS%7E1%5CTemp%5Cmsohtml1%5C01%5Cclip_filelist.xml"><style>
<!--
/* Font Definitions */
@font-face
{font-family:宋体;
panose-1:2 1 6 0 3 1 1 1 1 1;
mso-font-alt:SimSun;
mso-font-charset:134;
mso-generic-font-family:auto;
mso-font-pitch:variable;
mso-font-signature:3 135135232 16 0 262145 0;}
@font-face
{font-family:"\@宋体";
panose-1:2 1 6 0 3 1 1 1 1 1;
mso-font-charset:134;
mso-generic-font-family:auto;
mso-font-pitch:variable;
mso-font-signature:3 135135232 16 0 262145 0;}
/* Style Definitions */
p.MsoNormal, li.MsoNormal, div.MsoNormal
{mso-style-parent:"";
margin:0cm;
margin-bottom:.0001pt;
text-align:justify;
text-justify:inter-ideograph;
mso-pagination:none;
font-size:10.5pt;
mso-bidi-font-size:12.0pt;
font-family:"Times New Roman";
mso-fareast-font-family:宋体;
mso-font-kerning:1.0pt;}
/* Page Definitions */
@page
{mso-page-border-surround-header:no;
mso-page-border-surround-footer:no;}
@page Section1
{size:595.3pt 841.9pt;
margin:72.0pt 90.0pt 72.0pt 90.0pt;
mso-header-margin:42.55pt;
mso-footer-margin:49.6pt;
mso-paper-source:0;
layout-grid:15.6pt;}
div.Section1
{page:Section1;}
-->
</style>
<p class="MsoNormal"><span lang="EN-US">Hi,<br></span></p><p class="MsoNormal"><span lang="EN-US">I did the benchmark on wired LAN,
and it shows the same characteristic as in the wireless LAN. The performance is
shown in the attachment. </span></p>
<p class="MsoNormal"><span lang="EN-US">I did not add CPU utilization options.</span></p>
<p class="MsoNormal"><span lang="EN-US">The send throughput and the receive
throughput is roughly the same. If I plot them in one figure, it will be hard
to distinguish them. </span></p>
<br>Peng<br><br><div class="gmail_quote">On Sat, Feb 21, 2009 at 2:19 AM, Rick Jones <span dir="ltr"><<a href="mailto:rick.jones2@hp.com">rick.jones2@hp.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
<div class="Ih2E3d">Andrew Gallatin wrote:<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
Peng Wang wrote:<br>
<br>
<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
The benchmark platform I use contains two buffalo wireless g router. One<br>
router is the sender, and one is the receiver. The operating system in<br>
them is Linux with 2.4.7 version kernel.<br>
<br>
The result of the benchmark confuses me. From the following figure we<br>
can see that the throughput is highest when the UDP packet size is about<br>
20k bytes which is much bigger than the MTU. I guess it is because of<br>
the buffer in the system, but I am not quite sure. Do you see it before?<br>
Could you explain it?<br>
</blockquote>
<br>
<br>
I'm not at all familiar with performance characteristics of<br>
wireless networks, but on fast (10GbE) wired networks I<br>
sometimes see similar behavior when the host has expensive<br>
system calls. The larger the message size, then the more data<br>
is written on the socket using a single system call.<br>
</blockquote>
<br></div>
It doesn't necessarily have to be an "expensive" system call either. When sending 1000 bytes at a time that is 20 system calls and 20 calls through the UDP code into IP and then 20 trips through the driver and onto the "wire." With 20K bytes at a time, that is one system call, one call through UDP into IP and then 20000/MTU trips through the driver. If MTU is larger than 1000 bytes then that will be rather less than 20 trips through the driver - perhaps only about 15 or 16.<div class="Ih2E3d">
<br>
<br>
<blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, 204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;">
So, a few questions:<br>
<br>
1) Does a test via a 100Mb or 1Gb connection between the routers show<br>
the same characteristic? How about a local (loopback) test<br>
from a machine to itself?<br>
<br>
2) Have you measured the system call overhead using something like<br>
lmbench?<br>
</blockquote>
<br></div>
I would ask:<br>
<br>
3) Where are the netperf-measured service demands? (ie add CPU utilization options to your test)<br>
<br>
4) Does the graph show reported send throughput, or actual receive throughput?<br>
<br>
5) Was send throughput == receive throughput?<br><font color="#888888">
<br>
rick jones<br>
</font></blockquote></div><br>