[sclug] Newbie, partitioning 120Gb HDD - recommendations?
Alex Butcher
lug at assursys.co.uk
Tue Jan 18 11:50:08 UTC 2005
On Tue, 18 Jan 2005, Tom Dawes-Gamble wrote:
> On Tue, 2005-01-18 at 10:42 +0000, David Newcomb wrote:
>> Dickon Hood said:
>>> I also agree with someone at sun.com (I forget who; it's quite irritating)
>>> about disc partitioning: don't bother. One big / slice is easier to
>>> manage, and isn't much of a problem if you're careful.
>>
>> I also agree with this. It also means there is only 1 partition table
>> held in memory (and being worked on). If a machine has 4 or 5 partitions
>> then there will be a performance overhead with that.
I would be /extremely/ surprised if either of those statements are confirmed
by a read through of the kernel source code.
The only inherent performance penalty I would expect would be at boot, when
the partition table is read and the start/end blocks for each partition are
figured out. This is particularly the case for logical partitions which are
stored as a linked list. In the overall scheme of things, though, such a
performance penalty will be virtually unnoticeable.
>> I tried this a couple of years ago and there is a marked improvement in
>> having a single '/' partition.
Depending on the tests performed, this I can believe. Of course, if the
tests only model situations that don't occur in normal operation, they're
irrelevant; either fix reality to match your partitioning scheme (which can
include rewriting bad code), or fix your partitioning to match your actual
usage patterns.
>> David
>
> Do you have a description of the tests and the times?
>
> Personally I don't see how more partitions as such make much difference.
>
> Lets say you have a disk and divide it in to four equal partitions. For
> this example we will assume the disk a has the same number of sectors on
> the inner cylinder and the outer cylinder. Copying a file from
> partition one will be quicker to partition two than it would be to
> partition four due to the longer seek time. However, Writing or
> Reading to any one of the four partitions should be a uniform speed.
I agree, modulo the decreasing transfer rate from rotational discs as you
get closer to the hub (the usual virtual->phsyical mapping used by hard
discs, though this is not guaranteed). In general, putting active partitions
(e.g. /tmp, /var/cache, /home, /usr) near the beginning of the disc gives
better performance than putting them at the end (a good place for
/var/spool, though).
> regards,
> Tom.
Best Regards,
Alex.
--
Alex Butcher Brainbench MVP for Internet Security: www.brainbench.com
Bristol, UK Need reliable and secure network systems?
PGP/GnuPG ID:0x271fd950 <http://www.assursys.com/>
More information about the Sclug
mailing list