Re: Raw I/O vs files (was Re: Raw partitions / cooked files)

In article <2506@coyote.UUCP> gainer@slowmo.almaden.ibm.com (P. Gainer) writes: >How many DBAs are running Oracle/Informix/etc etc using raw I/O? Any? I suppose small db's use files and large db's usually use raw devices. I don't have figures on performance. But it is clear that a properly implemented db kernel can use raw devices better because it *knows* the data. (Oracle V5 was very poorly implemented in this sense). For small db's I don't think there is much difference in performance between file vs. raw. For large db's (with large buffer caches) I think there is some difference. Of course it depends on many things. I have always (last 5 years) used raw devices. The reason is not performance but convenience, and it makes sense only for large databases. In current setup we have 14 disks (1.7 GB seagates) dedicated to the db. I split each disk in 4 pieces and use the pieces (mirrored) as db files. It's simpler than maintaining many mirrored filesystems and db files in those file systems. I also feel safer with raw devices: a careless rm or a file system error can't delete the db files. - peno@kps.se