One solution, assuming you are already caching connections (i.e. mod_perl or some such), is to use the temporary, in-memory tables in SQL::Statement (available to any of the DBDs which subcalss it - AnyData, CSV, File, DBM, etc.). Basically you can then query the cache with the same SQL you use to query the full db and you can switch back and forth between them. Pair Networks, our kind host, uses this kind of caching for some of their databases and Dan Wright, their lead software developer will be talking about that and AnyData at the upcoming YAPC in Toronto. As for updating, the in-memory tables are updateable, so you can go in either direction (refresh the cahce from the db or refresh the db from the cache).