Data Movement

Every distributed computation ultimately executes as a local computation—a single computing process. You must move every piece of input data across the network from wherever it resides to the machine that needs to process it, and every piece of output data must travel over the network from the machine that produced it to its ultimate destination. Additionally, you can use caching to optimize data movement, providing a strategy for lowering the amount of data transfer. Moving large amounts of data over a network efficiently is a crucial aspect in the design of most distributed applications. Efficient data movement can often make a dramatic difference in performance.