A cluster deployment in a VMware environment involves in-guest SFW installation.

The following diagram represents in-guest SFW installation with support for VMware vMotion. The application data is stored on the shared disks and the cache area is created on the locally attached SmartDisks.

In the above example, SFW and the application that is configured for high availability is installed on the virtual machines (VM1 and VM2). A VMware cluster is configured between the ESX hosts. Multiple SSDs are locally attached to the individual ESX hosts. Using VxDMP (version 7.1 or later), these SSDs are aggregated to form individual SmartPools. From a SmartPool, multiple SmartDisks are provisioned for caching. A SmartDisk is exposed to a virtual machine as a RAW LUN (Here, SmartDisk1 is attached to VM1 and SmartDisk2 is attached to VM2).

For details about creating and administering SmartPools, and provisioning the SmartDisks, refer to Dynamic Multi-Pathing Administrator's Guide

After vMotion, the first I/O to the cache area fails and the request is forwarded to the back-end storage (data disks). SFW considers the I/O failure as a vMotion event and informs the VxDMP layer to allocate a SmartDisk of the required capacity. After the VxDMP layer allocates the SmartDisk, SFW retrieves the cache area details from the registry and creates a similar cache area on the SmartDisk. This newly created cache area is assigned to the virtual machine after vMotion.