Subject: Re: Varying disk geometry
To: None <tech-kern@NetBSD.ORG>
From: der Mouse <mouse@Collatz.McRCIM.McGill.EDU>
Date: 11/24/1995 10:05:02
>> Anyone here work for a disk manufacturer?
> Why, is the team soliciting bribes?
I have no idea; I (who posted the line you replied to) certainly am
not. I was actually hoping to get info from someone who knew whether I
was right or was spouting nonsense. As it happened, I did get such a
reply; it seems my guesses used to be basically right but are wrong for
modern drives. Seek time is pretty unpredictable and cannot be modeled
as I described, except perhaps piecewise, and many drives have onboard
caches that mean the time to fetch a block depends on whether you hit
the cache, which for our purposes is unpredictable.
I think we can just about forget all attempts at doing anything with
drive geometry and just treat a disk as a big array of blocks, with
performance bonuses for locality of reference, and let it go at that.