<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=us-ascii">
<meta name="Generator" content="Microsoft Exchange Server">
<!-- converted from rtf -->
<style><!-- .EmailQuote { margin-left: 1pt; padding-left: 4pt; border-left: #800000 2px solid; } --></style>
</head>
<body>
<font face="Calibri" size="2"><span style="font-size:11pt;">
<div>Hi,</div>
<div>I am performing a bandwidth comparison study on CORBA using Infiniband and 10GeT link.</div>
<div>I created a simple CORBA app to test the CORBA performance on IB vs 10GeT. </div>
<div>A CORBA servant object is hosted on Windows R2 server with 64GB RAM and Intel Xeon 2.5GHz processor.</div>
<div>The clients are hosted in RHEL server with 36GB RAM and Intel Xeon 2.5GHz processor.</div>
<div> </div>
<div>There have been two variables which were used. Each remote call is repeated until the total data sent is 8GB</div>
<div style="text-indent:-18pt;padding-left:36pt;">a)<font face="Times New Roman" size="1"><span style="font-size:7pt;"> </span></font>Remote calls are made with different argument sizes from 1kB,2kB,4kB upto 1mB.</div>
<div style="text-indent:-18pt;padding-left:36pt;">b)<font face="Times New Roman" size="1"><span style="font-size:7pt;"> </span></font>Number of client threads is varied from 1 to 32(Each thread splitting totalData/NumOfThreads)</div>
<div> </div>
<div>The surprising result is that CORBA performance on 10GeT is better than Infiniband.</div>
<table width="637" style="width:382.5pt;">
<col width="70" style="width:42.3pt;">
<col width="80" style="width:48pt;">
<col width="158" style="width:95pt;">
<col width="164" style="width:98.6pt;">
<col width="164" style="width:98.6pt;">
<tr height="24" style="height:14.4pt;">
<td style="background-color:#C6EFCE;"><font color="#006100"><span style="background-color:#C6EFCE;"> </span></font></td>
<td style="background-color:#C6EFCE;"><font color="#006100"><span style="background-color:#C6EFCE;"> </span></font></td>
<td style="background-color:#C6EFCE;"><font color="#006100"><span style="background-color:#C6EFCE;">IB(theoretical max: 54Gbps)</span></font></td>
<td style="background-color:#C6EFCE;"><font color="#006100"><span style="background-color:#C6EFCE;">10 GeT(theoretical max: 10Gbps)</span></font></td>
<td style="background-color:#C6EFCE;"><font color="#006100"><span style="background-color:#C6EFCE;">1 GeT(theoretical max: 1Gbps)</span></font></td>
</tr>
<tr height="24" style="height:14.4pt;">
<td style="background-color:#FFCC99;"><font color="#3F3F76"><span style="background-color:#FFCC99;">CORBA</span></font></td>
<td> </td>
<td>B/w:<span style="background-color:yellow;">8.66</span></td>
<td>B/w:<span style="background-color:yellow;">8.76</span></td>
<td>B/w:<span style="background-color:yellow;">0.864</span></td>
</tr>
<tr height="24" style="height:14.4pt;">
<td style="background-color:#FFCC99;"><font color="#3F3F76"><span style="background-color:#FFCC99;"> </span></font></td>
<td> </td>
<td>MTU:2044</td>
<td>MTU:9048</td>
<td>MTU:63</td>
</tr>
<tr height="24" style="height:14.4pt;">
<td style="background-color:#FFCC99;"><font color="#3F3F76"><span style="background-color:#FFCC99;"> </span></font></td>
<td> </td>
<td>Parallel:26</td>
<td>Parallel:32</td>
<td>Parallel:32</td>
</tr>
<tr height="24" style="height:14.4pt;">
<td style="background-color:#FFCC99;"><font color="#3F3F76"><span style="background-color:#FFCC99;"> </span></font></td>
<td> </td>
<td>KB/call:256</td>
<td>kB/Call:256</td>
<td>kB/Call:64</td>
</tr>
<tr height="24" style="height:14.4pt;">
<td style="background-color:#FFCC99;"><font color="#3F3F76"><span style="background-color:#FFCC99;"> </span></font></td>
<td> </td>
<td>Window scaling :ON</td>
<td>Window scaling :ON</td>
<td>Window scaling :ON</td>
</tr>
<tr height="24" style="height:14.4pt;">
<td style="background-color:#FFCC99;"><font color="#3F3F76"><span style="background-color:#FFCC99;"> </span></font></td>
<td> </td>
<td> </td>
<td> </td>
<td> </td>
</tr>
<tr height="24" style="height:14.4pt;">
<td style="background-color:#FFCC99;"><font color="#3F3F76"><span style="background-color:#FFCC99;">Socket</span></font></td>
<td> </td>
<td>B/w<span style="background-color:yellow;">:43.6</span></td>
<td>B/w:<span style="background-color:yellow;">9.35</span></td>
<td>B/w:<span style="background-color:yellow;">0.951</span></td>
</tr>
<tr height="24" style="height:14.4pt;">
<td style="background-color:#FFCC99;"><font color="#3F3F76"><span style="background-color:#FFCC99;"> </span></font></td>
<td> </td>
<td>Rec.Window:1M</td>
<td>Rec.Window:512K</td>
<td>Rec.Window:63K</td>
</tr>
<tr height="24" style="height:14.4pt;">
<td style="background-color:#FFCC99;"><font color="#3F3F76"><span style="background-color:#FFCC99;"> </span></font></td>
<td> </td>
<td>Send.Window:256</td>
<td>Send.Window:256K</td>
<td>Send.Window:256K</td>
</tr>
<tr height="24" style="height:14.4pt;">
<td style="background-color:#FFCC99;"><font color="#3F3F76"><span style="background-color:#FFCC99;"> </span></font></td>
<td> </td>
<td>Parallel:6</td>
<td>Parallel:2</td>
<td>Parallel:3</td>
</tr>
</table>
<div> </div>
<div>Here Socket performance means the bandwidth measured by iperf.</div>
<div> </div>
<div>With this results, I have following questions.</div>
<ol style="margin:0;padding-left:36pt;">
<li>I read in several old articles in internet that ORB kills the throughput because of inefficient memory copies and marshalling/unmarshalling. In 10Gb Ethernet and 1 Gb Ethernet, I do see almost 90% utilization(CORBA compared to ethernet). But IB card give
less than 25% of network utilization. What is causing this overhead?</li><li>Though CORBA is meant to be used for remote method calls, we are amazed with its superiority in serializing. Can we use it as alternative for socket data transfer with amnual serializing code? Is CORBA suitable in high bandwidth networks for data transfer?</li></ol>
<div> </div>
<div>Thanks</div>
<div>Krishna</div>
<div> </div>
</span></font>
</body>
</html>