Messages in this thread | | | Date | Sat, 20 Mar 1999 13:54:23 -0800 (PST) | From | Tim Smith <> | Subject | Re: [OFFTOPIC] Re: disk head scheduling |
| |
On 20 Mar 1999, Linus Torvalds wrote: > And I bet this will be even more true in the future. So I don't believe > it is worth doing today, and I certainly don't believe it will be worth > puruing in the long run. > > That's not to say that the current algorithm in Linux is perfect. It > isn't. There are probably interesting things that can be done with it. > But taking the disk geometry into account is not one of those > interesting things, I strongly suspect.
Probably true, but for the sake of completeness, it *is* possible, by careful and extensive observation and probing, to determine either the real drive geometry, or something that is functionally equivalent in the sense that if you use it for seek and rotational scheduling, you get correct results.
Linus is almost certainly right that its not worth it--I sure don't want to put up with around 12 hours of probing at install time to get the real geometry...and probably another several hours of probing whenever the disk detects and remaps a bad block.
There's an interesting paper by, I think, the same people that did the soft update stuff in BSD that describes their work in this area. They also have available a very interesting disk simulator that you can give the real geometry of a given disk, and it does a very good job of simulating the access time on that disk. They've got some graphs in their paper that show simulated response to various I/O sequeunces, and measured response on the real disk, and the graphs coincide almost perfectly.
--Tim Smith
- 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/
| |