Abstract: Disclosed is a storage efficient multi-key sharding approach for use in a data management system. The data residing on the data management system can be partitioned based on multiple columns independently of one another, thus providing faster querying on each of the columns used for partitioning. The more partitioning attributes other than the key attribute of data records in the data records being partitioned can be used. Specifically, a multi-key partitioning approach consistent with embodiments described herein can be applied to perform load balancing between multiple parallel servers or server processes.