In the event of disaster recovery SRM doesn't need both sites to be active. It is not like fault tolerance in that it doesn't keep a permanent replica fired up and ready in case of an outage, with minimal (if any noticeable) downtime.
SRM failover will invoke any protected VM's in the primary site to power up on the replica versions that are running on the secondary site. It is usually advertised as "one click recovery" which will perform all fail over tasks for you (if setup correctly). It is also possible to automate fail over entirely but that becomes a great deal more complicated.
The fail over process of SRM relies on storage replication under the hood, be it using VMware Replication Appliance or third party vendor storage replication. Depending on the type of replication and solution used, you get a a choice of RPO and RTO's to fit with your organizations business requirements for disaster recovery. Generally the nearer the RPO and the quicker the RTO, the more you pay!