A couple of things, I always try and place swap on a dedicated disk or a disk containing rarely used filesystems. Placing swap on the same disk as /var or any other logging locations is not a good idea. If you must cram everything on one disk, put swap on first, which will place the data closer to the spindle and reduce access times.
As for IDE vs. SCSI, for starters, IDE requires cycles from the CPU to perform it's function, SCSI does not. SCSI disks also usually have much better seek times, more IOPS, and usually spin faster (this is a constant per drive spec, not variable). (SCSI is up to 15k rpm, IDE is still at 7200rpm as far as I know) The difference between 7200k and 10000k is enormous, and could very well look like an order of magnitude faster but I believe it is still linear.
10k rpm SCSI drives are nearly 2-3 times as expensive as the same capacity in an IDE disk, but the performance gain is WELL worth it. Plus the lifetime of a SCSI drive will always be longer, on average, than IDE drives.
I got hooked on SCSI 5 years ago and have never looked back.
Hope this helps,
-Chuck
Luis Hernandez-Garcia wrote:
Hi, We've noticed in our lab that the machines that have IDE hard disks run their write operations(ie- things that use a lot of swap space) much more slowly than when we execute the same sort of operations in WIndows machines (Matlab, for example), and orders of magnitude slower than the machines that have SCSI drives. The IDE machines grind almost to a halt, whereas the SCSI ones fly through the operations, and the Windows+IDE machines do just fine.I know that SCSI is inherently faster than IDE and the Linux/Unix have been developed with SCSI in mind. But, isn't there a way to optimize the system so that it uses the IDE drives more efficiently? Any hints? I, and everyone around me would sure appreciate them. -Luis
-- redhat-list mailing list unsubscribe mailto:redhat-list-request@;redhat.com?subject=unsubscribe https://listman.redhat.com/mailman/listinfo/redhat-list
