Sequences of operations that end-users execute in the are rendered as commands in . When a job is executed, the recipe of steps is executed against the entire dataset. In the following diagram, you can review the layers that execute recipes across your dataset.
Platform execution layers
- recipe: End-users create these sequential recipes by performing point-and-click operations on sampled data in the application. These recipes are stored in the and can be executed at any time against the full dataset.
- Dataflow Compiler: This component compiles your recipe into a set of commands that can be executed against the specified execution target.
- Dataflow Optimizer: The optimizer reviews the set of compiled commands to perform general optimizations for the queries.
- Target Optimization and Specialization: The compiler performs additional optimizations and addresses any special requirements depending on the execution target.
- Extensible Function Library: Optionally, developers can create user-defined functions using Java.
- Execution Environments: The supports a variety of environments for execution of jobs.