Partitioning a network graph into smaller subgraphs (blocks) to enable localized computation and communication.