Dell PowerVault MD3260i - Manual (Page 97)

Synchronous Write Mode
Synchronous write mode provides the highest level security for full data recovery from the secondary storage array in
the event of a disaster. Synchronous write mode does reduce host I/O performance. When this write mode is selected,
host write requests are written to the primary virtual disk and then copied to the secondary virtual disk. After the host
write request has been written to the primary virtual disk and the data has been successfully copied to the secondary
virtual disk, the RAID controller module removes the log record on the replication repository virtual disk. The RAID
controller module then sends an I/O completion indication back to the host system. Synchronous write mode is selected
as the default value and is the recommended write mode.
Asynchronous Write Mode
Asynchronous write mode offers faster host I/O performance but does not guarantee that a copy operation has
successfully completed before processing the next write request. When you use Asynchronous write mode, host write
requests are written to the primary virtual disk. The RAID controller module then sends an “I/O complete” indication
back to the host system, without acknowledging that the data has been successfully copied to the secondary (remote)
storage array.
When using Asynchronous write mode, write requests are not guaranteed to be completed in the same order on the
secondary virtual disk as they are on the primary virtual disk. If the order of write requests is not retained, data on the
secondary virtual disk might become inconsistent with the data on the primary virtual disk. This event could jeopardize
any attempt to recover data if a disaster occurs on the primary storage array.
Write Consistency Mode
When multiple replication relationships exist on a single storage array and have been configured to use Asynchronous
write mode and to preserve consistent write order, they are considered to be an interdependent group that is in the
Write consistency mode. The data on the secondary, remote storage array cannot be considered fully synchronized until
all of the remote replications that are in the Write consistency mode are synchronized.
If one replication relationship in the group becomes unsynchronized, all of the replication relationships in the group
become unsynchronized. Any write activity to the remote, secondary storage arrays is prevented to protect the
consistency of the remote data set.
Link Interruptions Or Secondary Virtual Disk Errors
When processing write requests, the primary RAID controller module might be able to write to the primary virtual disk,
but a link interruption might prevent communication with the remote (secondary) RAID controller module.
In this case, the remote write operation cannot be completed to the secondary virtual disk, and the primary virtual disk
and the secondary virtual disk are no longer correctly replicated. The primary RAID controller module transitions the
replicated pair into an Unsynchronized state and sends an I/O completion to the primary host. The primary host can
continue to write to the primary virtual disk, but remote writes do not take place.
When communication is restored between the RAID controller module owner of the primary virtual disk and the RAID
controller module owner of the secondary virtual disk, a resynchronization takes place. This resynchronization happens
automatically, or it must be started manually, depending on which write mode you chose when setting up the replication
relationship. During the resynchronization, only the blocks of data that have changed on the primary virtual disk during
the link interruption are copied to the secondary virtual disk. After the resynchronization starts, the replicated pair
transitions from an Unsynchronized status to a Synchronization in Progress status.
The primary RAID controller module also marks the replicated pair as unsynchronized when a virtual disk error on the
secondary side prevents the remote write from completing. For example, an offline secondary virtual disk or a failed
secondary virtual disk can cause the remote replication to become unsynchronized. When the virtual disk error is
corrected (the secondary virtual disk is placed online or recovered to an Optimal status), then synchronization is
required. The replicated pair then transitions to a Synchronization in Progress status.
97