LVS
lvs-users
Google
 
Web LinuxVirtualServer.org

Re: Extremely slow director on Centos

To: "LinuxVirtualServer.org users mailing list." <lvs-users@xxxxxxxxxxxxxxxxxxxxxx>
Subject: Re: Extremely slow director on Centos
From: Somsak Sriprayoonsakul <somsaks@xxxxxxxxx>
Date: Mon, 06 Feb 2006 20:40:17 +0700
Thank you very much for the explanation! That really clear the fog for me.

Anyways, I still have to enable the firewall rule since the customer will not be pleased if they acknowledge the absent of firewall. Somehow the connection/second is quite impressive already. Almost no overhead for 4 real servers.

Roberto Nibali wrote:
The problem has been solved. It's something related to iptables.

As expected, it's netfilter and the connection tracking. If you want
high performance load balancing, do _not_ use netfilter; especially the
connection tracking. It just does not scale. Simply loading ip_conntrack
into the kernel makes your packet rate drop by 60 kpps on a 1Gbit/s
connection.

Stopping iptables on director and the connection rate goes from 200 to
Nx2000, where N is the number of real server.

Very well, so LVS works well for you.

After that, I tried to
figure out which iptables rules conflict with ipvs and found that it's
default argument generated from system-config-securitylevel that cause
this. Replace "-m state --state NEW -m tcp -p tcp --dport 80" with just
"-m tcp -p tcp --dport 80" make everything works perfectly.

It's not a conflict, it's the connection tracking core which is
extremely slow. There's ongoing effort from the netfilter people to
improve this state.

Regards,
Roberto Nibali, ratz


--
-----------------------------------------------------------------------------------
Somsak Sriprayoonsakul

Scalable Computing Lab
High Performance Computing and Networking Center
Kasetsart University
ssy@xxxxxxxxxxxxxxxxxx
-----------------------------------------------------------------------------------


<Prev in Thread] Current Thread [Next in Thread>