<html>
<head>
<style>
.hmmessage P
{
margin:0px;
padding:0px
}
body.hmmessage
{
FONT-SIZE: 10pt;
FONT-FAMILY:Tahoma
}
</style>
</head>
<body class='hmmessage'>Thanks Brian. Will do<BR><BR><BR><BR>
<HR id=stopSpelling>
> Date: Wed, 14 May 2008 09:53:18 -0500<BR>> From: brian.stansberry@redhat.com<BR>> To: wagad@hotmail.com<BR>> CC: jboss-cluster-dev@lists.jboss.org<BR>> Subject: Re: [jboss-cluster-dev] TCP vs. UDP<BR>> <BR>> Please post on the JBoss Clustering User Forum at <BR>> http://www.jboss.com/index.html?module=bb&op=viewforum&f=64 . This <BR>> mailing list is for discussions related to the development of clustering <BR>> features in JBoss projects, not for usage questions.<BR>> <BR>> M Wagad wrote:<BR>> > We have 2 node JBOSS Cluster, and we are having performance issue, what <BR>> > we found out is that if we turn of the slave server, and operate the one <BR>> > master node (Changing the load balance to 100/0), we have much better <BR>> > performance then when we have both of the running sharing the load with <BR>> > 50/50 load balance.<BR>> > <BR>> > Would changing the default Protocol of UDP to TCP make any difference in <BR>> > terms of Performance?<BR>> > <BR>> > I had come across these posts on the Net that praises TCP over UDP for <BR>> > organizations that have less then 16 JBOSS Nodes :-<BR>> > <BR>> > <BR>> > <BR>> > <BR>> > For TCP…<BR>> > <BR>> > <BR>> > <BR>> > /“We experiment with another configuration using TCP/IP and show that <BR>> > current J2EE application server clusters up to 16 nodes (the largest <BR>> > configuration we tested) can scale much better with this configuration. /<BR>> > <BR>> > /We attribute the superiority of TCP/IP based group communications over <BR>> > UDP/IP multicast to a better flow control management and a better usage <BR>> > of the network switches available in cluster environments.”/<BR>> > <BR>> > <BR>> > <BR>> > And UDP..<BR>> > <BR>> > <BR>> > <BR>> > “Having to support reliable multicast UDP communication is not a piece <BR>> > of cake. Drop a couple of packets and you get only a fraction of the <BR>> > throughput. Then lose the ordering of the packets and you are in a <BR>> > slump. The situation can turn pretty sour if you use large messages that <BR>> > get fragmented (more than one UDP packet). You only need to lose one <BR>> > packet to lose your entire message - then you resend everything. The <BR>> > point is that in unideal and congested conditions you start to miss TCP, <BR>> > and it is probably extremely hard to beat its adaptive protocol that has <BR>> > been perfected for so many years.”<BR>> > <BR>> > <BR>> > <BR>> > <BR>> > <BR>> > I would also like to know if any companies out there make the change <BR>> > from UDP to TCP and if they can share their experience with us.<BR>> > <BR>> > <BR>> > <BR>> > Thanks<BR>> > M. Wagad.<BR>> > <BR>> > <BR>> > ------------------------------------------------------------------------<BR>> > <BR>> > <BR>> > ------------------------------------------------------------------------<BR>> > <BR>> > _______________________________________________<BR>> > jboss-cluster-dev mailing list<BR>> > jboss-cluster-dev@lists.jboss.org<BR>> > https://lists.jboss.org/mailman/listinfo/jboss-cluster-dev<BR>> <BR>> -- <BR>> Brian Stansberry<BR>> Lead, AS Clustering<BR>> JBoss, a division of Red Hat<BR>> brian.stansberry@redhat.com<BR><br /><hr /> <a href='' target='_new'></a></body>
</html>