On Wed, 8 Apr 2009, netadmin @dslextreme.com wrote: > I setup a computer has two 1GB interfaces, doing NAT mode, > when sending traffic to it from Load Runner to the server, > I saw the traffic going up and down between 1000 to 1500 > transactions per second, not very stable. If I just use > the same computer as router
as a NAT router? > the traffic chart is much flat and get about 2000 > transactions per second. > I also tested the max concurrent connection using IXAX > tester, it limited to about 3 million and no matter how > much memory on the sytem, it will not go up. The new > connection per second rate is also about 6k/s. and if you change the director to a NAT router what numbers do you get? > I also noticed 2.4 kernel LVS works better than 2.6 kernel > on the same hardware, performance wise. that's interesting to know. How much better? I remember a long time ago someone staying with a 2.0.36 kernel because it ran LVS much faster. > From those tests, I think there are some limit in > somewhere, especially the trsanaction rate with and > without LVS having that much differences. It's good to have data. Most people don't run tests like you have and send us the results. Much of the current code was developed on machines running at 100-400MHz. Much of the effort is put into making sure that LVS is not the rate limiting step in the flow of packets through a director. However the code is not put through a battery of speed tests. Joe -- Joseph Mack NA3T EME(B,D), FM05lw North Carolina jmack (at) wm7d (dot) net - azimuthal equidistant map generator at http://www.wm7d.net/azproj.shtml Homepage http://www.austintek.com/ It's GNU/Linux! _______________________________________________ Please read the documentation before posting - it's available at: http://www.linuxvirtualserver.org/ LinuxVirtualServer.org mailing list - [email protected] Send requests to [email protected] or go to http://lists.graemef.net/mailman/listinfo/lvs-users
