PostgreSQL has had memcached support for a VERY long time
pgfoundry.org
pgfoundry.org
(FWIW, AFAICT, the main reason MySQL needed this is because its prepared statements functionality is subpar. In specific, it does not seem to support preparing a parameterized statement and then saving the query plan to be used for a later bound set of arguments: when the arguments are later bound it is forced to re-optimize the query. I imagine that most of the same performance benefits of HandlerSocket could then easily be obtained using the PostgreSQL v3 protocol, albeit with obviously different client support.)
http://use-the-index-luke.com/blog/2011-07-16/planning-for-r...