The number of nodes in an HBase cluster is typically driven by physical size of the data set and read/write throughput.