On planet.d.o I find Russell Cokers blog one of the most interesting blogs at all. Today he blogged about swap sizes in the past and nowadays.
I agree with him that having 16 GB swap space on a 8 GB system is next to useless, but I disagree with this:
A Linux machine with 16M of RAM and 32M of swap MIGHT work acceptably for some applications (although when I was running Linux machines with 16M of RAM I found that if swap use exceeded about 16M then the machine became so slow that a reboot was often needed).
I think this is not always true. Instead it all depends on your swap setup, even more bigger machines (2-8 GB RAM). As some might know, I'm running some m68ks as autobuilders. All of these Amigas are quite low on RAM compared to today standards: 64 MB and 128 MB. On the other hand, some packages are require lots of memory during the build, sometimes even 700-800 MB (rare cases), which means lots and lots of swapping.
The trick is (of course) to distribute the swap space to several disks instead of just one big swap area. This gives a great performance boost while swapping, especially when swapping is not linear. Using 2 SCSI host adapters and 7 SCSI disks in one of the m68k made that machine very responsive even when using 300-400 MB of swap.
So, distributing the swap space to several disks (with same swap priority) gives a performance boost. But keep in mind that it might crash your machine, when one of the disk dies. Alas, this happens of course as well, when you use just a single swap partition. ;)