In order for Site Recovery Manager to failover virtual environments, the persistence layer of the environments needs to be readied at the target site. The persistence layer refers to the storage-which is protected by array-based replication. The process of "readying" is typically called a failover. A failover is usually comprised of a few steps, synchronizing the data on the target side with the source side, making the source side inaccessible, making the target side accessible, and vendor-specific operations that surround those operations.
Since all vendors implement their replication differently (and have correspondingly heterogeneous APIs) VMware implemented a translation layer for SRM that provides a common set of APIs for communication across all vendors. This layer is implemented as a plugin by the storage vendor and within it is where vendor-specific operations are executed. This translation layer is called a Storage Replication Adapter--usually abbreviated as an SRA. The SRA receives SRM specific API calls (synchronize a data set for instance) and then issues that to the array in a way that the array understands.
The vendor creates the SRA, validates it via a comprehensive test suite, then makes the SRA available to customers for download. SRM does not come packaged with these SRAs, so a customer must download it then install it into the SRM environment.
SRM comes in two varieties, a Windows-based implementation, and a Linux-based appliance implementation. The windows based implementation has been deprecated for many years now though, so the linux version should be the one in use. The process to install the SRA is slightly different for each version so the process for each will be outlined on this page.