----- Original Message -----
From: "June G. Gonzales" <[EMAIL PROTECTED]>
To: "'The Main Philippine Linux Users' Group (PLUG) Discussion List'"
<[email protected]>
Sent: Thursday, September 29, 2005 9:30 PM
Subject: [plug] good partitioning practice
Hi,
What should I observe when partitioning a HD?
What are the correct sizes for each partition?
Is there a certain percentage that I should a lot for a specific partition
on HD of XX Gig.
For example I have a 20 Gig HD.
basically you need at least 5 partitions and these are:
1. /
2. /var
3. /tmp
4. /usr
5. swap
the size of each partition depends on what you are trying to do or what
services you are planning to run on it...
the root "/" partition is the catch all partition once there is a directory
that is not being partition... the root partition usually in a fix size and
doesnt grow its disk usage in order to protect from disk full because this
is the most sensitive partition of all partitions...
/var partition is where usually the logs, print spoolers, mailboxes, etc are
stored... it is partition to avoid disrupting the entire system when there
is a misconfigured or filling the entire /var partition....
/tmp partition is where usually programs creating a temporary files..... it
is partition to avoid also disrupting the entire system due to disk full
also when programs were not properly clean its temporary files upon exit or
abnormal termination...
/usr partition is where usually programs, user's home directories and future
addtional programs are stored... this is the biggest disk space allocation
of all partitions (usually the rest of disk space)...
swap partition is where usually paging and swapping are using it.... the two
are the same by moving the process to and from the disk but differ on the
amount of move.... swapping is moving the entire process to and from the
disk while paging is only moving page by page.... the size of swap
partitions depends on your ram and the services you are running....
basically, if you have small amount of ram you have to allocate more swap
space because once your ram is filled up... it is started swapping those
idle processes to swap space and give more ram to the active and agressive
ram hunger process... if you have more ram and those processes are not idle
all the time... you can allocate less swap space... one good example where
you have huge ram but you allocate more swap space where you have lots of
students remote access your server and idleing most of the time... it is
worth noting that once you see there is swap space used... it clearly
indicates that you are running of memory unless there is a process that was
doing a memory leak.. a memory leak where a process always allocating a
memory and forgot to free it for the rest of its running time...
if you notice that i didnt mention any actual partition size.. it is up to
you to experiment what size is best fit for that particular partition for a
given system...
furthermore, some notes for you to ponder with....
1. you can create a /boot partition that allocated below 1024 cylinder if
you have older bios that cannot read partition above 1024 cylinder.. this
partition is where your kernel file stored for boot up...
2. outer cylinder is faster to access both for read and write compare to the
inner cylinder (track 0)... therefore move your heavy disk i/o partition to
the outer cylinder and less disk i/o partition to the inner cylinder..
3. if you have multiple physical disk... spread your swap partition across
it... if you can stripe it (eg. raid 0, 0 + 1) move your other partition
there..
4. if you forgot to partition something for example /home and needs bigger
space because it belongs to root "/" partition... use softlink to your
bigger partition.. eg. ln -s /usr/home /home ..... same true with /var/tmp
where you softlink it to /tmp partition... take note that linux has a bug on
stat(2) function call on symbolic links and i dont know if they are already
fix that bug...
5. every applications have different requirement on its disk i/o speed....
for example, proxy cache average file is 11kb while databases are bigger..
try to experiment and tune the partition blocksize, bytes per inode and
cylinders per group for maximum performance...
6. use the proper filesystem... for example, if you have terabytes
storage.... you have no where to go but to use journaling filesystem...
under this filesystem, the recovery time is pretty much faster than the
non-journaled filesystem when trouble comes...
fooler.
_________________________________________________
Philippine Linux Users' Group (PLUG) Mailing List
[email protected] (#PLUG @ irc.free.net.ph)
Read the Guidelines: http://linux.org.ph/lists
Searchable Archives: http://archives.free.net.ph
