Elevator vs first-come-first-served
Robert Minichino (rmini@joni.pasture.net)
Mon, 16 Nov 1998 06:16:00 -0500 (EST)
The modified elevator algorithm as currently used is more than adequate for most
workstation and light server use; it's reasonably fair, and most workstation
disk drives do little to no reordering (buffers on typical IDE drives range from
64-512k). However, with an intelligent caching disk controller (either on the
HBA, or a SCSI-SCSI RAID controller), the sorting of disk blocks can actually
decrease performance. Granted, I haven't poked around in the Linux disk code
too much recently, but I wonder how difficult it would be to make either a
kernel or runtime option to disable the elevator algorithm and instead use
simple first-come first-served seek optimization. This way the intelligent
controllers that -do- know the physical geometry of the disks can do their job,
and we can still accelerate disk accesses on systems lacking this luxury (most
of them). Just my US$0.02.
--
Robert Minichino
Chief Engineer
Denarius Enterprises, Inc.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/