Running enterprise applications like Microsoft Exchange 2013 on vSphere might require a different operational approach. For instance, a dedicated vSphere “island” cluster can be configured with a different configuration than typical vSphere configurations, such as including rules to avoid over-commitment, limiting vSphere HA / DRS / vSphere vMotion, or dedicating storage resources to performance-intensive mailbox virtual machines.

The concept of island clusters is fairly simple. An island cluster (also referred to as a dedicated application cluster) hosts workloads with special license, performance, availability, or other configuration requirements. Some software vendors apply licensing policies on their applications, middleware, and databases that are not conducive to virtualization, and especially DRS, where the application can potentially touch a high number of physical CPUs. Island clusters are one approach to dealing with this challenge.

Some service providers also use island clusters of operating systems such as Windows or RHEL. This helps them save money on data center Windows socket licenses, which are typically the most cost effective way of licensing large numbers of Windows VMs running on a host. Another benefit of this approach is that it helps ESXi take advantage of the memory management technique of Transparent Page Sharing (TPS), which is now disabled by default by VMware. While this can be more efficient due to the higher chances of duplicate pages being spawned by VMs in physical memory when multiple instances of the same guest OS virtual machines are running, this approach is not recommended for island clusters supporting virtualized business critical applications. Some use cases for island clusters are represented in the following figure.