Re: Elevator vs first-come-first-served

Robert Minichino (rmini@joni.pasture.net)
Mon, 16 Nov 1998 14:13:57 -0500 (EST)


> Given the general range of opinions on this subject, and especially
> the possibility that the "usual" sorting might reduce performance, it seems
> to me that the best option is to modularize.
...
> 2) A loadable module (per system? SCSI controller? disk drive?
> filesystem? depends upon the hardware) could be used to provide
> a more advanced sorting algorithm.

But I don't see a clean way to interface such a module in the system at
this time. Needs hacking. Oh well :)

> If you want, after the system boots, a user-level program
> could measure the read timing on the drives (after the rest of the
> system has settled down? during the initial startup itself?
> interesting options to explore) and load the sorting module(s) with
> configuration-specific parameters.

Or a program that's runnable once and generates timing information for
each block on the drive. It would take a while to run, but it would be
interesting to see whether or not the benefits would be substantial. A
userland program could then feed it to the sorting module. Yet another
research project.

> It would be an interesting student research project, for
> example, to make a neural net request sorter...

Or even a relatively simple expert system sorter would do. This would also
be a good tool to demonstrate the possibility of starvation (shortest seek
time first) when sorting blocks. All in all, I think this will be a Good
Thing.

--
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/