Within its particular sweet spot (in-RAM data structures) I think it is probably the best option out there. I also like the fact that it seems to be moving in the direction of becoming "memcache++"; providing a useful superset of interesting data structure operations in addition to the standard set of memcache features that most people have come to rely upon. It almost seems that memcache is now so tightly coupled to the standard RDBMS mindset that there is nowhere else for it to go. This does not diminish its utility, but it makes something like Redis more appealing by offering the same set of features plus a few nice additions for some other common data types.
That's definitely what has me interested; we have the standard MySQL + memcache (plus svn, to make things interesting) stack, and the idea of of nonvolatile "shortcut" around MySQL built into our cacheing system is very appealing.
