I have always created my ADABAS databases on RAW disk (for a time that was my only option on UNIX). I’m in the process of installing a new server and will be upgrading from ADABAS 3.3 to 5.1 and the question has come up again… Does ADABAS perform better on RAW disk or File System (COOKED)?
What is the opinion of other Open Systems ADABAS DBAs?
Does ADABAS perform better on RAW disk or File System (COOKED)?
To be honest: This is an interesting question, but we never did a proper test. We got a lot of ADABAS-Databases, but they are all smaller than 6 GB and the number of users is never bigger than 10. So this question of performance never came up.
There are another questions to think about:
If you use “Cooked”, it’s easier move the Database to anywhere else in the filesystem. You just have to change some configuration files.
It’s easier to add a new container on a filesystem without booting the system.
I imagine that when you first installed adabas on raw disks, the performance delivered by these devices was nowhere near the performance delivered by more modern disk packs (probably with built in caching). The management of an adabas database in filesystem is easier than in raw, so, imho, if the performance is adequate from the file system, then go with that, otherwise, use raw.