Storage Systems - SQL Server 2012 Internals and Troubleshooting - page 89

Because the server is tied to a storage network, the different tiers of disks don't even need to exist
within the same array. Bob could place his data, log, and tempdb i les on one array and move his
backup data to a completely different storage system. This approach offers the added benei t of
shrinking the possible failure domain. If the primary array suffers a catastrophic failure, the backup
data still exists on a separate array.
This mixing and matching of storage within a SAN array is called storage tiering. Some storage
arrays provide automated storage tiering , monitoring volumes or pieces of volumes for high perfor-
mance. When predetermined performance characteristics are detected, the array will migrate data
onto a higher tier of storage. Using this model, Bob only needs to put all his data on a storage vol-
ume; the array determines where to place the data and when to move it.
Different storage vendors have implemented storage tiering in unique ways. One of the most unique
features is the granularity of data that is migrated. Some arrays use a transfer size of a few kilo-
bytes. Other arrays migrate gigabytes of data. Each system offers different performance. Be sure to
evaluate any potential system to ensure it meets your needs.
WARNING Storage tiering has the effect of changing performance over time.
Many i nancial systems always require consistent performance. If your specii c
system requires repeatable performance then storage tiering should not be used.
Data Replication
SAN arrays offer both internal and external storage data replication. Internal replication consists
of data snapshots and clones. Some storage arrays offer inner array data migration features. Both a
snapshot (also called a snap) and a clone offer a point-in-time data copy. This data copy can be used
for backup or reporting.
NOTE For more information on increasing the performance of reporting
systems please visit: http://sqlvelocity.typepad.com/blog/2010/09/
scalable-shared-data-base-part-1.html .
Both snapshots and clones need to be created in sync with the database log and data i les. In order
to maintain SQL data integrity, both the log i le and data i les need to be copied at exactly the same
time. If the log and data i les are not in sync, the database can be rendered unrecoverable.
Prior to creating the point-in-time data copy, you need to decide on the type of SQL Server recov-
ery that is needed. You can create application-consistent copies using the SQL Server Virtual
Backup Device Interface (VDI). VDI is an application programming interface specii cation that
coordinates the freezing of new write operations, the l ushing of dirty memory buffer pages to disk
(thus ensuring that the log and data base i les are consistent), and the fracturing of the clone or