Case study 10: HP-UX: to raw devices or to file system: that is the question!
Platform: HP-UNIX 11
Sybase ASE version: 12.5.x
Background story: We encountered terrible performances when moved our database from Windows testing environment to HP-UX production environment. Switching to HP-UX showed some 'black holes' in our system which could not be seen at Windows. After some changes has been made in front-end, I expected major changes in HP-UX performance tests, but results have been still significantly better on Windows then on HP-UX. There are other things which could be discussed about different performances on Windows and HP-UX, but here I want to describe what improves HP-UX performances greatly.
Task: Find out what would be the benefits from switching to raw devices instead of devices on file system.
Solution: Switching to raw devices showed significantly better performances. More about tests which proves this can be found in this .doc file
These tests are conducted in controlled environment, without other users. Only difference in these tests was the thing I want to measure: raw devices and file system.
Apart from using standard sp_sysmon Sybase procedures for performance monitoring I find several useful HP-UX commands for monitoring OS performances during critical tests. One of them is sar command. More about sar can be found by using
man sar
There are graphical tools which can interpret sar results. One of them can be found on this site.
Other useful commands fro HP-UX are iostat and vstat.
