[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Partition



On Fri, Nov 17, 2000 at 02:18:27PM -0500, Brian Pontz wrote:
> Whats the reasoning behind having multiple partitions on the same hard
> drive?

cd ~
dd if=/dev/zero of=manisthisbig

Wait, and then see what happens.  Or, read:

http://www.openbsd.org/faq/install-i386.html#1.4

> Is it for performance reasons?

Yes. Besides fragmentation, as listed in the link above, the first blocks on
drives are physically located toward the outside of the platters, where the
linear speed of the data vs. the heads is highest. That means more data per
rotation, which equals higher thoughput. By keeping high IO data toward the
outside, you get better trhoughput.  By placing heavily used data in partitions
located toward the outside edge of the platters, you ensure the maximum 
throughput for that data.

Also, if you can keep heavily/often accessed data in close physical proximity
on the platters, you can decrease seek time, especially in the case of data
toward the outside edge of the platters. Since the outside edge holds more
data, the number of tracks you need to seek past from one block to another is
decreased. 

On my servers I tend to order it as:

/
swap
/var
/usr
/home

because most of my high IO work is done on in /var.  (/ and swap take up
relatively little space, so having them at the edge of the platters doesn't
waste much, and in the case of swap, is bennefical.)

YMMV.  I used to have a series of Bonnie benchmark resuls comparing two
equal size partitions; one at the begining and one at the end of the disk.
As I recall, there was at least a 30% throughput difference, and roughly the
same diffenece in seek times.

-Paul

-- 
/Paul M. Hirsch              /
/elektrosatan@voltagenoir.org/
/GPGPGPkeyID: 0xD11A250E     /