After changing in its benchmarking policy, and while waiting the launch of VMmark, a complete benchmarking system for virtualization platforms, after years VMware starts to disclose some performance measurements on its products:
For this set of experiments, we compared a baseline running multiple WebSphere Application Servers on a single server to running each application server on a VMware virtual machine (VM) in a single server. This allowed for the simplest comparison configurations. Each server was separately administered, and each application executed independently, with no load balancing performed across the applications. A shared, non-virtualized, database was used by all applications.
For the baseline measurements in this paper, we chose to run multiple applications using separate application server instances. Running multiple application servers works well when a single server does not use the maximum performance of the server, that is, it cannot maximally utilize the CPU, memory, and other resources.
Another approach is to use virtualization to run multiple guests machines, each with a single instance of the application server. This approach was used for the VMware measurements in this paper. This approach provides greater isolation between the applications. For performance comparison purposes, these experiments ran WebSphere Application Server instances in each guest machine, however, virtualization also provides an opportunity to consolidate different types of applications…
Read the whole whitepaper at source.
Thanks to Run-Virtual for the news.