Roughly ordered based on priority.
- Workflow language: Provide simple Python library, that allows users to generate workflows by just writing code and feeding it to a parser.
- Visualization: Provide a simple visualization tool, that allows users to see the status of the workflows and See the execution visually of a workflow invocation.
- Error handling in workflows: Exception Handling, allowing users to deal with errors in functions. For example, users could provide a fallback function that is executed when the first function fails.
- Observability: Add initial telemetry support, measuring function runtime, workflow engine overhead.
- Performance optimizations: Pre-warm functions, the main optimization that would make the workflow engine faster than calling functions yourself. The workflow engine interprets the dependency graph, notices that a certain function will be called ‘soon’, and triggers the specialization of the function by Fission before it has to call it.
- Function versioning: provide support for dealing with versions of functions. Users can indicate how to workflow should deal with new versions. Examples of strategies for dealing with versioning: never upgrade, canary deployment, blue/green deployment...
- GUI: Build a user-friendly GUI.
- Data/Control flow split: Add support for passing around data ‘by reference’. Currently, all data passes through. the workflow engine, which can be costly for data-intensive tasks. Support can be added to allow functions to pass data to each other directly.
- Catalog: Create an online ‘catalog’ of functions, such that users can re-use functions created by others.
- Non-functional requirements: Improve scheduling decission
- Benchmark: Create benchmark that measures the performance of Fission Workflow.