Can file placement improve my disk performance?
Extensive laboratory tests on this subject have produced mixed results.
Some controlled-environment tests have shown that if the INDEXF.SYS file is placed in the middle of the disk, and the directory files are placed near that, and the most commonly used files are placed near that, disk access time can be shortened.
While this all sounds good at a theoretical level, it simply does not hold up in a live, multi-user, multi-tasking, production environment.
I/O requests from all user processes are intermingled. The files that are frequently accessed by one user may not be the same as those frequently accessed by others. The frequency of access may change at different times of the day or times of the month. In many cases, the most frequently accessed files are very large master files or databases, but the frequent access is actually to a small number of the blocks in those large files.
Other factors such as directory caching, disk geometry, and bad block re-vectoring further dilute the real-world application of the theory that file placement is valuable.
The real answer to reducing I/O access time is twofold:
-
Keep files defragmented, so that reading them does not require split I/Os.
-
Instead of trying to place frequently-accessed files on the disk to shorten access time, keep frequently-accessed blocks in memory where they can be accessed with no disk I/O at all!
dynamic data caching system from Executive Software does this. It adjusts to changing system resources and user requirements, automatically.
If this FAQ helped, please let us know. If you didn't find the answer to your question here please e-mail
|
