Splitting the Home dir two different partitions

The 3 primary reasons you would create these as separate partitions are as follows:

1. performance
2. isolation
3. security

Examples

• By separating /home you can put this data on a shared network disk so that when UserA logs into servers in a given domain, their /home/$USER will be a single copy that follows them from machine to machine. This is typically done using NFS and automounts (aka. autofs).

• By putting /usr data on it's own partition, it can be mounted read-only, offering a level of protection to the data under this directory so that it cannot be tampered with so easily.

• Some additional reasoning for isolating /usr, is for making it easier to deploy identical systems, these partitions can be prepared one time and then replicated across systems more easily.

• Also separating the data out can make it easier for backup cycles.

• Finally separating volatile directories such as /home can protect a system from having it's primary disk fill up by either an accidental or malicious user.

Over the course of my 15+ years of doing this I've only ever seen /home separated (as a network share via NFS) and the /boot and /var directories as being isolated as separate partitions. Outside of some esoteric Solaris boxes I can't recall ever seeing a Linux system having a separate /usr - and note that if you don't have /usr mounted before init starts, your system will break in esoteric and silent ways.

Martin Monperrus wrote a wonderfully quick guide to auto-mounting on Linux using the udev package. Here's a copy of the ussefull bits but be sure to check the comments there too

sudo nano /etc/udev/rules.d
# automounting usb flash drives 
# umask is used to allow every user to write on the stick
# we use --sync in order to enable physical removing of mounted memory sticks -- this is OK for fat-based sticks 
# I don't automount sda since in my system this is the internal hard drive # depending on your hardware config, usb sticks might be other devices than sdb* 
ACTION=="add",KERNEL=="sdb*", RUN+="/usr/bin/pmount --sync --umask 000 %k" 
ACTION=="remove", KERNEL=="sdb*", RUN+="/usr/bin/pumount %k" 
ACTION=="add",KERNEL=="sdc*", RUN+="/usr/bin/pmount --sync --umask 000 %k" 
ACTION=="remove", KERNEL=="sdc*", RUN+="/usr/bin/pumount %k"

Note the bits about KERNEL=="sdb*", and KERNEL=="sdc*", are relotive and you may wish to use UUID's to make sure that udev auto-mounts the correct drive to the correct directory path regardless of what USB port it gets re-plugged into. A quick search pulled up an Arch Linux forum post on using uuid instead of sdb* or sdc* error prone way of mounting. Once the drive auto-mounts properly then it should only require adding the proper export PATH=/mnt/media/some_usb/program/:$PATH line to the shell's enviroment variables so that the program can be called by name for that user.