Nowadays the whole dedicating limiting volume sizes seems to be going to the wayside as there is always eventually a case where you run out of size due to unforeseen changes in how things are stored or what is stored, so it's easier from the management point of view to just create huge arrays and let them grow and extend them as needed.
Things that were typical worries addressed by physically limiting volumes...swap space...log files spilling over and crashing the system...etc. seem to be getting addressed by cheaper hard disks, filesystems that are no longer limited to smaller sizes than available drive sizes, and volume management that lets you dynamically resize and add/subtract disks as needed. Oh, and I remember having to worry about drive failures too, but now these virtual volumes can be masking RAID volumes under the filesystems. Performance is more a matter of drive spindles and the application of the server (heavy writes? Heavier on reads? Need equal performance?) than a simple stick in a drive and share it out solution.
The only real drawback we've seen is that large volumes can take a really long time to perform disk checks on, but it's normally not a huge problem with journaling filesystems in use.
Our normal routine is to create a system partition for the OS, then throw everything else into a giant data partition for shares, home directories, etc. Usually by the time we outgrow it (performance or space reasons) we need to replace the server. Other admins on the site with experience dealing in larger multi-terabyte SANS and such might have other experiences to share.