Database RAID Array Planning

A database RAID array (redundant array of independent drives) can provide large amounts of input and output (I/O) throughput and capacity, while appearing to the operating system and RDBMS as a single large disk (or multiple disks, as desired, for manageability). The use of RAIDs can greatly simplify the database layout process by providing an abstraction layer above the physical disks, while promoting high performance.

Performance of the RAID feature provided by the operating system might not be satisfactory. To obtain the best RAID performance, use the RAID support provided by your RAID vendor.

If a RAID Array Is Not Used

If a RAID device is not in use, even if space is at a premium, then you must separate indexes with names ending in _P1 from the tables on which they are created. These tables are heavily used in joins.

If you make frequent use of Siebel Enterprise Integration Manager (EIM), then you might want to put the EIM tables and indexes (names starting with EIM_) on different devices from the Siebel base tables. Both tables are accessed simultaneously during Siebel EIM operations.

Microsoft SQL Server RAID Array Planning

The following table describes a sample disk layout for a server dedicated to Microsoft SQL Server, where the database uses a single filegroup residing on a disk array. The use of a single RAID array for the database devices provides satisfactory performance in many cases without the administrative overhead of using individual filegroups.

Disk	Objects	Comments
Single mirrored	Windows OS	N/A
Single disk	Windows pagefile	Segregate for maximum performance.
Single mirrored	SQL Server logfile	Segregate sequential I/O for database performance.
3 to 5 disks (minimum) in a RAID configuration	Siebel database data and indexes	Add as many spindles as required for performance and storage capacity.

If your enterprise requires the highest performance standards, then place heavily used tables and their corresponding indexes, such as those listed under Sizing the Database for a Siebel CRM Deployment, in a specific SQL Server filegroup within your database. By creating a filegroup on a specific disk or on multiple disks, you can control where tables and indexes in your database are physically located. For more information, see Database Physical Device Planning.

When separating database objects into filegroups, you can avoid complex calculations by using Microsoft’s recommended RAID disk layouts.

Your choice to use RAID devices or multiple filegroups to distribute database objects depends solely on how great your performance needs are. It is recommended that you work with your hardware vendor to determine the optimal RAID configuration for your specific requirements.