May 4, 2009
    At the TeamQuest Technology Summit, John Miecielica of Metavante gave an informative talk on how to combine performance management and capacity planning with monitoring. He explained how much easier it used to be to do performance testing when you only had one server attached to its own storage with a defined set of users. Nowadays with�virtualization, it's a lot tougher as multiple users and apps are sharing a common infrastructure.

    He gave an interesting example of an application that was suddenly exhibiting performance issues, yet had no change in traffic patterns and was running the same workloads. The application and infrastructure staff insisted no changes had been made.�Performance monitoring indicated a change in IO and how it was serviced, yet the SAN team said there had been no microcode changes, new apps or major changes in storage subsystem.�The issue had been caused by another team that had loaded an application on a shared server.

    They had been very happy with the success of their implementation and had no idea of the consequences.

    Yet a heavily virtualized environment makes such events inevitable. It takes tools like TeamQuest to provide enough visibility to detect and prevent similar occurrences.