You can improve system performance by placing logs and database objects on separate devices:
Placing a table on one hard disk and nonclustered indexes on another ensures that physical reads and writes are faster, since the work is split between two disk drives.
Splitting large tables across two disks can improve performance, particularly for multi-user applications.
When log and data share devices, user log cache buffering of transaction log records is disabled.
Partitioning provides multiple insertion points for a heap table, adds a degree of parallelism to systems configured to perform parallel query processing, and makes it possible to distribute a table’s I/O across multiple database devices.
See Chapter 6, “Controlling Physical Data Placement,” in Performance and Tuning: Basics for a detailed discussion of how object placement affects performance.
