[lastcanal]

The major advantage (depending on your use case) is that your dataset is not limited to the size of your available memory.

[otterley]

Does it manage its own cache like InnoDB? Otherwise you're gonna have a bad time.

Redis tried this (see http://redis.io/topics/virtual-memory) but it died a quick and well-deserved death (the code isn't even in it anymore as of 2.6).

[lastcanal]

It uses LevelDB as the storage engine and lets you configure the cache parameters from the configuration file.

   leveldb:
	# in MB
	cache_size: 500
	# in KB
	block_size: 32
	# in MB
	write_buffer_size: 64
	# in MB
	compaction_speed: 1000
	# yes|no
	compression: no