Detailed analysis of Spark Shuffle process
There are some differences in the process of shuffle between shuffle and Spark in MapReduce. Here is a specific introduction.
Diagram of shuffle process of Mapreduce
Spark shuffle process diagram
Note: there are no partitioning and sorting procedures in the spark shuffle process, and the stored results are stored in memory, so it is much faster than mapreduce.
That's all for now. The illustration should be clearer. If you have any questions, please leave a message.