Yes, the bottleneck is the disk I/O, it seems. But what to do, it is already in RAID10 15k disks. I could add some more spindles.
The vendor is just testing the app for performance. First we should go to v8 from 2000i SP4. Then convert the dbs from v5 to v8. If the db is compact, disk I/O-s should be less.
Then comes the memory question. 2000i can use only 640 MB for cache. v8 can use the whole preset %. Is this dynamic cache that intelligent, that if I give 4 GB more then it can cache the most used db pieces for read/write, and therefore reduces that much disk I/O?
The question is that from the nearly 500 smaller db files and 1 5,5 GB, 1 4 GB, 2 2,5 GB, 1 2 GB dbs (in 2 GB chunks) could it put it in the dynamic cache and runs the server almost from RAM?
How can I test it out, how much memory could it effectively use?
PerfMon?
Thanks mirtheil