hi ya walter
good question ...
- anybody measued any of the partition schemes ??
am assuming "a" is the outer tracks .... "d" is the inner tracks...
-- swap should be NOT an issue... if you sustain constant 32Mb or 64Mb or
128Mb of swap space usage .... ADD that much more memory !!!
-- you should optimize for os calls and data ...
- what is the server doing most of the time...
- calculating something or read/writing to disks ??
i tend to use the following partitions... ( just because ... no particular
reason as far as disk geometry
hda1 / 64Mb to allow single user boots to fix thesystem
hda2 /tmp 128Mb
hda3 /var 512Mb or whatver for web/email servers
hda5 /usr 4Gb for system styff
hda6 swap 2x memory
hda7 /home rest of disk
so i suspect... most of the head is moving between the first few
4Gb of disk...and as you add more and more data under /home.... you'd
stay
more towards the center of the disks ??
am NOt sure you can optimize your partitions for general server issues...
but if you know you are gonna be computing pie or doing finite element
analysis.... thats a whole new ball game ???
- how much time and $$$ does one wanna spend to get computations
done in 1 day vs 2 days or 1 week vs 2 weeks..etc..etc..
-- put all user data on a separate disk would be a good start
-- using (properly) stripped disk to allow reading 2x as fast is good...
-- using scsi disks might be a better disks for high thruput disk transfers...
c ya
alvin
On Fri, 25 May 2001, Walter L. Preuninger II wrote:
> I wouild like to optimize my partitioning scheme
>
> Example A
>
> space a b swap(center of drive)
> c d
>
> Example B
>
> a b swap c d
>
> Example C
>
> a b c d
> swap(center of drive)
example C does NOT seem to be center of disk as compared to example A
> In this example, i assume a to be /usr and c to be /var
>
> Which example gives the best performance.
>
> Does 2.2 do elevator seeking
>
> So the root question is, should sawp be between the most active file
> systems, or should swap be in the center of the drive
>
> Thanks,
>
> Walter
>
>
> --
> To UNSUBSCRIBE, email to [EMAIL PROTECTED]
> with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
>
