Open hanxuanliang opened 4 months ago
The current Kis-flow does not yet support DAG (Directed Acyclic Graph) scheduling, but we can perform parallel streaming operations.
For example, in cases like combining flow1 and flow2 to generate flow3:
We can have FunctionS and FunctionL nodes, where FunctionS acts as a storage node and FunctionL serves as a read/write node. These nodes can link the two flows together. Alternatively, we can have flow1's result sent to a message queue, and then flow2 can use this as a data source, initiating another flow to execute.
For configurations like:
pipeline:
- depend_on: flow1
like the above config
This configuration format is not yet supported. However, it's a very good suggestion. The approaches mentioned for Kis-flow can be adapted to a configuration-based approach.
Maybe DAG is a little more common. It may also be necessary to introduce the docker runtime environment
Will the kis-flow support DAG scheduling later?
like the above config