CouchDB metadata is about 120 bytes per document
Since we have a lot of documents to be indexed (1000 per second), we build up 10GB of couchDB metadata per bucket (not related to real data) which must be kept in memory *per day*. Originally we thought that small documents were in the confort zone of couchbase, but while memcached handles small documents well, couchDB does not.
This means that we need to manually compact documents in to a single one to reduce the amount of overhead on the couchbase for storage. Unfortunately while couchDB is doing fine with large documents, memcached is not, especially the 20MB limit means that if we want to go bigger t we get out of the spec of current couchDB. What would be the best way of getting the best of both worlds?