Table of ContentsView in Frames

SQL Server configurations supported with SnapManager

SnapManager databases can be configured on one or more storage systems. This section shows the various ways that you can place the data of your SQL Server on storage system volumes.

Note: If you change the database configuration after performing a SnapManager backup, you might not be able to perform an up-to-the-minute restore using that backup. Therefore, perform a backup immediately following any configuration changes.

SQL Server configuration requirements for SnapVault backups

If you archive a database to a SnapVault backup (clustered Data ONTAP only), the database and SnapInfo directory must be on separate volumes.

SQL Server configuration requirements for SMB shares

Multiple databases on different LUNs within the same volume

The following supported configuration shows multiple SQL Server databases sharing the same volume but residing on different LUNs.

Multiple databases on one LUN

The following illustration shows multiple SQL Server databases and all their associated files and transaction logs on one LUN.

This is a simple configuration, and it can be applied to an SQL Server that supports about 35 databases per volume.

Note: In this configuration, all databases in the shared LUN are backed up at the same time, even if certain databases have not been selected for the backup. However, you have the option to select which databases you want to restore from a multiple-database backup.

Multiple databases sharing two LUNs

The following illustration shows an example of multiple SQL Server databases and all their associated files and transaction logs sharing exactly two LUNs. The database files cannot reside on any other LUNs. The LUNs can be located on the same or different storage system volumes. The illustration shows an example in which each LUN is located on a different volume.

By placing the data files for multiple databases on one LUN and the transaction logs for those databases on the other LUN, SQL database performance is improved by separating the random I/O patterns of the data files from the sequential I/O patterns of the transaction log files.

Note: If you select to restore only a subset of the databases that reside on one or two LUNs shared by multiple databases, then a stream-based restore method is used rather than the online Snapshot restore method.

Single SQL Server and multiple storage system volumes

The following illustration shows a configuration in which the data and transaction log files of an SQL Server database reside on separate storage system volumes. Placing all transaction logs on one volume and using another volume for all the database files is partly due to performance. If the volume with the data files fails, it is still possible to back up the log file, restore the last full backup, and then apply all backed-up current transaction logs. This configuration requires another volume for the SnapInfo directory.

Multiple SQL Servers and one storage system volume

When the SQL Server environment does not generate high I/O load, a single volume can optimize the use of disk and volume space. However, this configuration has two disadvantages:

Multiple SQL Server instances on the same storage system volume

The following illustration shows a storage system volume with LUNs containing the data files of multiple SQL Server instances residing on a storage system volume that is different from the volume on which the LUNs for the transaction log files reside.

Note: Using three volumes prevents the loss of the complete SQL Server environment and makes it quicker to restore from unmanaged media if a volume goes offline.

Multiple file groups belonging to the same database on different LUNs

The following illustration shows multiple file groups belonging to the same database residing on different LUNs within the same storage system volume.

Multiple VMDKs belonging to the same database on different volumes

The following illustration shows multiple file groups belonging to the same database residing on different VMDKs on the same datastores.