mapreduce - Does CouchDB really split views across servers? -
i'm delving couchdb, , puzzled distribution of map-reduce computations in views. see lot of resources mentioning map-reduce inherently distributed, because can process 1 half of data on server a, other half on server b, , reduce both results. 1 example slide 16 of presentation:
http://www.slideshare.net/gabriele.lana/couchdb-vs-mongodb-2982288
this seems logical, but:
couchdb not seem provide api dispatching computations several servers. distribution appears provide replication of entire data set other servers (which then, assume, compute own view data).
couchdb uses b-tree store view data based on keys generated in map step of view algorithm, precludes appropriate partitioning of documents based on server should on.
so, couchdb distribute map-reduce computations @ all? or map-reduce property used merely cache values in b-tree nodes?
you looking bigcouch, enables couchdb cluster , uses distributed mapreduce.
Comments
Post a Comment