Automatic bandwidth scaling through the use of in-cluster cache instantiations while retaining atomic writes

sorry if this is the wrong category/component

To save you time, the problem is outlined in #1 of the first post and the solution/design itself is in the fourth post down starting at "I've put more thought into that feature idea,"

See: http://www.couchbase.com/forums/thread/handful-questions-hotspotsemit-atomicityview-updatesreduce-results-saving

Other relevant parts (there are various topics in that large thread) are: #1 in the third post. edit #2 and edit #3 in the seventh post. the 1st paragraph in the eleventh post.

the only problem i can see is that the memory i'm suggesting the neighbor uses to cache a document is already being used to hold the working set itself. can you shrink the size of the working set for a single particular node on the fly / on demand? perhaps persist the LRU document(s) to disk to free up room in the memory for a cache instantiation? since cache instantiations do not persist and are not replicated, they have to be treated differently than regular documents in the working set

i don't know couchbase internals at all. maybe what i'm suggesting is infeasible given the current design