Jun 01, 2015, 12:28 PM
We've set up a Network Load Balancer and a web farm on the virtual host. At its maximum capacity, the virtual cluster contains 4 identically configured web server nodes each having dedicated 4-core CPU running at 2 GHz and 16 GB of RAM. We've observed a reproducible improvement in the application performance that directly correlates with the number of active nodes participating in the load traffic distribution. CPU saturation and thread queue length counters correlate negatively with the number of active nodes.
At this point, I'm inclined to believe that the application demands more computational power to run on the server so the chief bottleneck appears to be the CPU. Here's a quick summary report of the latest performance testing against the virtual cluster:
At this point, I'm inclined to believe that the application demands more computational power to run on the server so the chief bottleneck appears to be the CPU. Here's a quick summary report of the latest performance testing against the virtual cluster: