Chapter 8

1. The BSP computer is an abstract computer model made up of a collection of potentially heterogeneous processors that are interconnected via a computer network. Processors can not only access their own local memory, but they can also use the network link to exchange data with other processors. In other words, the BSP computer is effectively a distributed memory computer that can perform computations in parallel.
2. The Single Program Multiple Data (SPMD) technique models distributed data processing tasks as a self-contained piece of software that runs on a single-core machine. The program receives a set of data as input, applies a processing function to it, and emits some output. Parallelism is then achieved by splitting the dataset into batches, launching multiple instances of the same program to process each batch in parallel, and combining the results.
3. A super-step is...