Which OS provides the _fastest_ PostgreSQL performance

toby We may actually agree about all this. It seems to me that some of the following may apply:

1.) Perhaps the drive does not give "write complete" signal until the data are actually on the disk platter. If so, though, I see little point in keeping write data in the drive cache. Although maybe there is. At least it frees up the (SCSI in my case) controller quicker so it can do something else, while waiting for the drive to do the write. Further, it could send the next data to the drive, or read from the drive, before the original write completes. It would require a rather integrated design between the device driver and the device's cache manager though, and while that is possible, I have difficulty imagining that the drives are that smart. I hope I am wrong and they are that smart (i.e., reply queuing).

2.) In the old days when I designed a dbms for UNIX (late 1970s), we wrote the journal file to magnetic tape, and at the time UNIX wrote to tape very fast; i.e., did not leave stuff in the RAM cache any longer than absolutely necessary because the machines I had (PDP11-47, PDP-11-70) did not have enough memory to maintain large caches.

-- .~. Jean-David Beyer Registered Linux User 85642. V PGP-Key: 9A2FC99A Registered Machine 241939. ^^-^^ 06:05:01 up 19 days, 8:33, 3 users, load average: 4.14, 4.17, 4.09