On Nov 12, 2007, at 6:15 , Peter J. Holzer wrote:
Ah, well in the general case, there's no processing to do for not found keys.I don't think that's true. If the key hasn't been found in memcached you often need to look in the database (or whatever it is you are caching).
I was referring to the client work. It's implied that a cache miss means that you need to go look somewhere else.
If you know of the cache misses earlier you can fire off the first database requests while you are still receiving cache hits from memcached. But I think this is possible with normal get requests, no change to multiget required.
You *can* do that, but I'd expect it to be significantly slower and more resource intensive than getting the cache results and doing a single lookup against your source with all of the missing keys.
Of course, one could contrive a pathological case, but I imagine it'd take quite fancy code to be faster at incrementally handling multiple source queries while processing memcached results and still be faster at rendering its final result than doing it in two steps.
-- Dustin Sallings
