White Papers

Additional resources
16 Dell EMC SC Series: Disaster Recovery for Microsoft SQL Server Using VMware Site Recovery Manager | CML1018
instantiated hourly for that volume. The length of time the replication will take depends on the size of the
snapshot, which is dictated by the rate of change on the volume, and the bandwidth between the source and
target systems. In the given example, assuming the replication bandwidth is sufficient, a one-hour RPO is
established for the applications and data on the volume. A majority of customers implement asynchronous
replication because it provides a good balance between cost and acceptable RPO.
4.2.2 High consistency synchronous
High consistency synchronous replication is also configured on a per-volume basis and the data on each
volume is replicated to a remote system. While data replicated asynchronously is sent in batch intervals,
synchronous replicates written data immediately when it is written at the source. In effect, each incoming write
I/O becomes a dual-write process whereby data is written at the destination system and the source system.
After the write is committed in both locations, a write acknowledgement is sent to the operating system and
application where the write originated. With this in mind, synchronous replication provides a clear benefit over
asynchronous: zero data loss in the event of a disaster, providing the initial replication is not out of date and
replication is not paused. However, synchronous replication is more expensive in terms of application latency
and site connectivity costs. A highly reliable replication link between sites is needed to accommodate the bulk
of data being replicated while minimizing application latency at the same time.
Although synchronous replication is constantly replicating data as it is ingested at the source, it should be
noted that as of Storage Center OS (SCOS) 6.3, frozen snapshots are also replicated to the remote site
automatically. This provides flexible recovery options in that the data at the remote site can be recovered
either from the active snapshot or from a frozen snapshot.
4.2.3 High availability synchronous
High availability synchronous replication is the same as high consistency synchronous replication except that
replication is allowed to fall behind in favor of source site application availability during periods of extreme site
to site latency or an unplanned remote site connectivity outage. Because replication is allowed to fall behind,
data loss could occur in the event of a disaster at that time.
4.2.4 Asynchronous active snapshot
SC Series asynchronous replication also has the ability to replicate a volume’s active snapshot. An active
snapshot contains newly written data or data that has been changed on a volume since the last frozen
snapshot was created. When the replication of active snapshot data is enabled on a per-volume basis, the
data is replicated to the remote system as it is written to the source volume on an immediate but best-effort
basis. Replicating the active snapshot is often called semi-synchronous. It is similar to synchronous
replication except there is not a guarantee that data was written to the remote system before it is
acknowledged on the source system. The two types are similar as long as the remote site bandwidth
connectivity can support the data rate of change being replicated. However, if it does not, it is allowed to
immediately fall behind and the result is no application latency penalty or predictable data loss, should a
disaster occur at that time. Replicating the active snapshot is a feature of asynchronous replication, meaning
frozen snapshots will be replicated as well.
4.2.5 Live Volume
Standard asynchronous or synchronous (either mode) replication types can be leveraged by VMware
vSphere SRM protection groups, recovery plans, and reprotection. Live Volume adds an abstraction layer to
the replication to allow mapping of an abstracted volume derived across two SC series arrays.