Re: Disk geommetry, was Re: Kernel Upgrade: Why?
On Wed, Apr 21, 1999 at 03:56:31PM -0700, debian-user list wrote:
> The HOWTO, and the person responding to the Mindcraft survey, both assert that the part of the disk furthest from the spindle is the fastest, and that one can use this to optimise performance. While I don't doubt the first part of the statement, I was under the impression that once you hit a drive's onboard logic, geometry is pretty much up for grabs these days, and attempting to put something on the "outer edge" of the disk is an excerise in self-delusion. The HOWTO is dated 1997, and the information therein may be older. Does anyone out there have some knowledge of current hard drive manufacturing?
It's true that you can't gather much from the sector and head numbers on
a modern drive, but I think common drives can be expected to access mostly
sequentially across the cylinders. I've always heard that the lower-numbered
cylinders were on the outer tracks. How many filesystems start packing
data in at the beginning? Would a drive manufacturer want to optimize or
decrease performance in that situation?
> Or better yet, some numbers from formal or informal experiments in drive partition performance?
Yes, I'd like to see this too. Wouldn't be hard to do, but I don't have a spare drive at the moment.
- Marsh
Reply to: