Big Data Computing

33Toward Evolving Knowledge Ecosystems for Big Data Understanding

Sharding is a client-side affair, that is, the database server does not do it

for user. In this kind of environment, when someone accesses data, the data

access layer uses consistent hashing to determine which machine in the clus-

ter a precise data should be written to (or read from). Enhancing capacity to

a sharded system is a process of manually rebalancing the data across the

cluster. The database system itself takes care of rebalancing the data and

guaranteeing that it is adequately replicated across the cluster. This is what

it means for a database to be horizontally scalable.

In many cases, constructing Big Data systems on premise provides better

data ow perform