Virtualisation is on everybody’s lips at the moment although many mainframe aficionados will happily tell you that it has been around for many decades albeit in a different guise.
Virtualisation has been designed to maximise hardware investment by using hardware resource more efficiently. In the past, it was common place to have one application hosted on one physical server. Statistics have shown that typically these servers used, on average, around 18% of their hardware resources, leaving 82% unused! Virtualisation is now used to re-cress this balance. Virtualisation technology now sits on top of the “pool of physical resources”, allowing the creation of multiple virtual servers which share these hardware resources. These virtual servers have no knowledge of their “neighbours” and believe they are, in fact, physical servers.
In this way, virtualised servers provide voluminous cost benefits but at the same time also increase risks by having many applications and services run in a centralized virtual environment. Any outages or performance degradation in one virtual server could result in many users on other virtual servers getting affected. This is due to the fact that if comprehensive management of the virtualised environment is not carried out 24x7, the shared resources could become scarce should unexpected performance issues happen on one of more virtual servers. This interdependency between virtual servers highlights a significant difference between physical and virtual server management and something IT management solutions should address.
Furthermore, there are 2 distinct levels of monitoring for virtualised environments, the top level monitoring from the perspective of the host server and the monitoring of individual virtual servers. The former level will monitor the health and performance of the host and overall health and performance of the virtual servers. However, to get a true picture of the health and performance of all server detailed monitoring of the individual virtual servers is required, including services, processes etc. Only this way will you get a level of monitoring comparable to physical server monitoring