Load balanced front end 1

[pjb]

Since moving our front end web server from a physical machine to a virtual using VMware, response has taken a big dive. Our server group is now suggesting adding a second front end virtual machine and using a Foundry Network Serverlon load balancer to flip back and forth. Does anyone have experience in this?

 

[Syntergy]

We have done quite a bit with performance tuning including setting up load balanced sites. There are two key settings to consider. The first is how the load balancer determines which front end Livelink instance to send a request to. It is hard to say there is preferred approach in a Livelink environment. It would seem that least connections would be the best approach but we have customers that have better performance using round robin. The issue is that some Livelink connections can be long running (workflow initiation or large file transfers) and other connections can be very short. The only way to be sure is to run some load simulations that mimic your Livelink usage and try to determine which is best for your environment.

The other key setting is how often the load balancer determines if a Livelink instance is alive. We have seen customers set the interval for checking to as low as once per second which affected performance quite a bit. We recommend about once a minute but again it depends on your Livelink environment. In this case there is a chance that some users will get sent to a non responding Livelink instance for up to a minute. This may not be acceptible in some organizations.

Syntergy offers performance tuning/analysis services including load testing if you happen to need any help in these areas.