Like Lindsay Lohan and Carly Rae Jepson, the data center without virtualization is slowly fading into the realm of memory alone. Unlike that pair, there are ways to monitor and control their behavior in the modern world. But in all seriousness, virtualization monitoring is not the same as monitoring physical devices and their applications in a hypervised environment.
The great value of virtualization in the data center is better resource utilization of CPUs, memory, and storage. When the business needs a new application or to give an existing one more (or less!) resources, it can be granted with little effort compared to provisioning a new server and installing it. In other words, change isn’t just good, it’s a sign of a healthy use of virtualization.
But there are 3 problems for monitoring these environments:
Keep monitoring current. The biggest danger here is not monitoring something that should be monitored. Typically, a new virtual instance is unmonitored until the monitoring solution is told about the new thing to monitor. This update can be done manually or automatically. GroundWork recommends the Lindsay Lohan/ automatic approach by having new things appear on scene, flash a smile to the biggest camera (i.e. the monitoring server) and saying “I’m here! Look at me!”
Monitoring at static thresholds. Here the danger is monitoring the wrong thing. When servers are reprovisioned or partitioned, their performance (and availability) may degrade before the monitoring solution can be told “This is OK; monitor this new configuration at these levels.” This can artificially make service level agreements (SLAs) and Operation Level Agreements (OLAs) look bad. Thresholds change as virtual resources change to match needs. This is good but must not be seen as down or slow time in your reporting.
Using separate tools. The major hypervisor vendors – Red Hat, VMware, Microsoft – make monitoring tools for their products. It’s natural to use them out of the box to ensure things are working properly. Often, they are not integrated into the overall monitoring setup though. Having too many tools can slow down access to the right contextual information to make quick decisions when needed. That context can usually be outside the virtualized environment, into networks, storage or even power/cooling.
Luckily, it’s not hard to get started with consolidated virtualization monitoring without laying landmines for the future. For starters, monitor just a few virtual servers at first – but fully; start with your vendor’s predefined Profile for a VM. Then increase the number until your entire virtualized environment is being monitored.
Based on our worldwide customer base, the most usage is in the United States and Europe, and is trending upward slowly in the rest of the world. Unlike your average attention-seeking Hollywood stars, your IT can be well-monitored without having to resort to ankle bracelets and paparazzi.