Ok, I've got 5 servers here. The application hitting them is extrememly memory intensive, taking ~340mb/user session connected. I wrote a load balancer for it 2 years ago based on memory only, and since there was plenty of memory, it worked flawlessly for that time, but users are growing and now machines are starting to page. They can handle it, but since they won't be getting a new server for a few weeks here, some of the servers have more processors than others, and thus is a bettery choice after the machine starts paging, but obviously only up to a point. So here's what we got server 1 - 28gb ram 4 cpu (started to page at 86 users) server 2 - 24gb ram 4 cpu (started to page at 75 users) server 3 - 16gb ram 2 cpu (started to page at 56 users) server 4 - 16gb ram 2 cpu (started to page at 56 users) server 5 - 16gb ram 2 cpu (started to page at 56 users) My original load balancer was just using memory, and we had plenty of room to grow, so memory was the only consideration. Now at 325 sessions, all memory is used and servers start paging, slowing the application a bit. The more cpu machines can take more load and reduce the amount of paging requried. Any ideas on how best to mathmatically load balance these things after they start paging?