Task is a basic computing unit of MaxCompute. SQL and MapReduce functions are completed by task.
For most tasks submitted by users, especially the computing tasks, such as SQL DML, MapReduce and so on, MaxCompute will analyze them and generate the task execution plan. The execution plan is composed of multiple execution stages which are dependent each other. Now, the execution plan is displayed as a directed acyclic graph; vertex in the graph designates the execution phase; while edges of graph indicate the dependence of each execution phase. MaxCompute will follow the dependency of execute plan to execute each pahse.
In an execution stage, there are muitiple processes, aslo called Worker, to complete the computting work. Different Workers deal with different data, but the execution logic is the same. Computational tasks will be instantiated. User can operate this instance. For example, Status Instance, Kill Instance and so on.
On the other hand, some MaxCompute tasks are not computational tasks, such as DDL statement in SQL. These tasks only read and modify the metadata information in MaxCompute. Therefore, no execution plan is needed.