I generally use Compress::LZF as "an extremely fast (not that much slower than a pure memcpy) compression algorithm. It is ideal for applications where you want to save some space but not at the cost of speed".
Google recently released Snappy, previously known as Zippy, which "aims for very high speeds and reasonable compression" and is used in Bigtable.
And luckily gray recently released Compress::Snappy.
Whenever slinging data around the network or writing things temporarily to disk, it costs you almost no CPU to use lightweight compression and will save you some I/O. Which is good as processors are fast and disk and network are slow.
For permanent storage I still use gzip, zlib or bzip2.