2 kind of workflows:
- 10 sequential long running tasks
- 40 sequential lightweight tasks
We took the Fibonacci numbers as a good template for both use case:
def fibo(n: int):
if n <= 1:
return n
else:
return fibo(n - 1) + fibo(n - 2)For the lightweight tasks we computed fibo(10) and for the long running tasks, we choose fibo(33).
We executed the benchmarks on an EC2 t2-medium instance, always using Docker compose to spin up the orchestrator.
Retrieving the results and assembling them is a very manual tasks and differs by orchestrator.