Golan Shem-Tov here with the solutions architects team at Quest.
We all understand the importance of monitoring OS when we monitoring application servers and Data Bases. A lot of us have seen the impact of running out of disk space or how another process hogging CPU causes problems for everything that was running on that OS.
With the introduction of virtualization we are starting to see the impact of virtual infrastructure resources in the field. Looking at an application without looking at the VM infrastructure one would think that this application is having problems that come from the application server itself.
Looking at the alarms you will see one coming from the application tier, chances are you may even see multiple ones for requests in different time frames having different problems (e.g. slowness, exception, incomplete).
Investigating these problems with traces to a method level will get you to a slow method and to the conclusion that you have code problems, but were we seeing the full picture?
Let’s look at the same problem with a small change – We add the virtual infrastructure to the definition of the application tier, it looks very different doesn’t it?
Looking at the alerts we can clearly see that the virtual machine that the application server is running on is not getting CPU cycles from its ESX server and causing performance issues.
Conclusion – Just like we monitor the OS that our applications and Data Bases are running on we have to monitor the Virtual infrastructure and add the Virtual Infrastructure monitoring to the application view or else we miss a lot and can potentially spend a lot of time investigating problems in the wrong place. Moreover, it is NOT sufficient to just monitor both the virtual infrastructure and the rest of the application – you have to be able to see the application service and its users IN CONTEXT of the infrastructure – be it physical or virtual. Content (data) from the virtual tier without the context (correlation) to the other tiers just adds another team to the widely-seen and problematic blame game.