There are several uses for cgroups. From the system administration probably the most important one is limiting resources -- the classical example here being the cpu access. If you create a group for e.g.
sshd and give it some non-negligible CPU time share (compared to other groups or the default under which fall all unsorted processes), you are guaranteed to being able to login even in times when the machine will be running a CPU-intensive tasks.
More interestingly, if you give this "remote access" processes much higher CPU share then the rest, you will be able to log in almost instantly (because the ssh daemon will be prioritised over the rest of running processes) while you won't hurt the overall computation strength of the machine, since the resources are allocated only on a per-need basis. You usually want to do this together with I/O (including network) prioritisation. However, as John correctly points out in the comment below, one doesn't want to do these things carelessly (since it might fire back in an unexpected way). Important thing to bear in mind is that the groups are inherited by default -- i.e. one doesn't want to start a memory/CPU hog from such an ssh session. Yet for this there are mechanisms that can assign processes to cgroups as they start.
Another use is isolating the processes from each other -- in combination with other features (namespace isolation) in recent Linux kernels they are used to create OS-level virtualization like the LXC (Linux Containers).
Other than that you can do various accounting and control stuff (freezing some processes groups, assigning them to specific CPU cores etc.).
The two links here, should be a reasonable starting place if you are looking for more information. You may also want to check
Documentation/cgroups directory in the Linux kernel source tree.