In-memory database scalability -

i have been exploring mmdb systems lately , haven't been able find information regards how in-memory database supposed scale. quite basic assumption main-memory db constrained memory available on db node, , operating system management of memory. how can expand in-memory system size beyond of main memory available? assume answer along lines of distributed system haven't got clear in head how work. , of course it's possible misunderstood idea of mmdb , i'm missing obvious.

a bit of background on question: writing number of cross-platform mobile apps (even though background heavily involved mysql , mongodb), , don't native database solutions sqlite android , ios. thought i'd write own solution (site , github) in javascript (i'm working on cordova/phonegap). realised make nodejs module , use db web app (i'm creating blog powered experiment , it's working pretty well), of course i'm thinking of making separate tier , started thinking obvious limitation of memory size, hence question.

in-memory databases scale in size same way on-disk (aka persistent) databases do: either throw more storage @ (memory, in case) or distribute across multiple nodes of cluster. latter alternative increases complexity (both of dbms, , administration of it), relative in-memory database on single system. consider difference between vanilla mysql , mysql cluster. and, you'll want have fast network times when dbms has perform inter-node operations (e.g. distribute data or pull data multiple nodes satisfy query).

there's nothing particularly special in-memory databases in regard. there special optimizations in database engine when know storage memory. doesn't change fundamental principles of database systems.

what don't want create in-memory database larger physical memory. you'll force os swap in-memory database pages in/out of swap space, , performance suck. you're better off, in case, using conventional dbms , giving cache have memory available for. dbms use cache more intelligently os' swap space.