The most popular implementation of clustering today is used to protect the availability of applications, such as Exchange and SQL, where two or more servers, acting as active/active or active/passive, share a commonly accessible data storage device.
The objective of the cluster being to ensure application availability in the event of hardware failure.
The design of the cluster is critical to avoid single points of failure – the resilience of the cluster is only as strong as the ‘weakest link’.
The closer the designer wishes to get to the “five 9’s” (99.999% uptime), the more numerous and expensive the components. New technology does make clustering a reality, but at what cost?
If the objective of the designer is to gain application availability, then a more effective and efficient solution can be built using the techniques of virtualisation.
The functionality of products like VMware protect from server failure. The management element, VirtualCenter, constantly monitors server hardware performance and availability.
In effect, VirtualCenter with its HA and DRS capabilities, manages and controls one big cluster or pool of hardware resources.
Through its management capabilities it can move and recover virtual servers from a failed physical server host to a healthy one.
When coupled with the current technologies available to replicate both data and virtual server across wan links the level of application continuity achieved goes beyond mere standalone clusters.