1. The kv data layout needs to be modify so that records are first ordered by bucket id and then by key. 4. The current algorithm for computing the kv partition id does not ensure that a given bucket id falls within a single kv partition.