Forum Discussion

Zmei_101634's avatar
Icon for Nimbostratus rankNimbostratus
Nov 10, 2011

LTM VE Trial performance problem




I am trying F5's Big IP LTM VE solution using the trial version (10.1.0) to use on my network.




The network configuration is quite simple:




the LTM with one IP, serving 4 Virtual Servers, each with only one real server.




The real servers do not have https, so 3 of the 4 Virtual Servers receive SSL connections from the clients, decrypt it, and send the request as HTTP. Trivial stuff.




I was impressed with the ease that I was alble to configure this, the UI is very friendly and intuitive.




But, When I run some tests the performance results were really bad.



I used Webserver Stress Tool 7 to test the responsiveness of the Big IP LTM and compared to the current loadbalancer I am using, an old Juniper DX, and agains the real server.



Avg response time from the juniper and real server was 4/5 ms over LAN.


Big IP's avg response time was around 8/9 ms over LAN.



I was targeting the one site that does not use SSL at all, all HTTP.




I tried multiple configurations, caching, no caching, compressing, no compression but the results didn't change.


Any ideas why the performance is so bad?



What am I missing?




Best Regards,















2 Replies

  • Hi,



    I may be speaking under correction, but you compared a VE against a hardware appliance? I would think that using something that runs on a hypervisor platform, compared to a physical device, would have some additional latency, that a hardware device would not have.



    Are you using any kind of TCP optimisation profiles for the client and server side connections on your virtual servers? Might help negate some of the latency.



  • Mike_Krasnow_10's avatar
    Historic F5 Account
    The Trial is not intended for any type of performance testing. I'd suggest you contact your sales team and get an evaluation key for the full version.