If I want to add scheduling to my Hamilton dataflows on a node basis (not just entire dataflows) … what are the best tried and tested options available for that? Effectively, I want my cached and materialized nodes, (that are going to be requested on-demand for analytics use in a frontend application) • logged every time they're executed • profiled for critical metrics (latency, timing, memory usage) detection • scheduled for preprocessing / soaking dynamically, based on intelligence gathered from profiling, and orchestrated/scheduled by a macro orchestrator / scheduler

decorators?

@Stefan Krawczyk yes, exactly. For the first two, I think that the mechanisms hamilton already provides, are sufficient. It is not clear to me though, if there's any well supported and well interoperating scheduler / macro orchestrator, … that might understand hamilton dataflows granularly enough to let me schedule arbitrary single nodes • without having to write that out in code manually as dr.execute([…]) selections of discrete nodes • and without having to basically write my own scheduler, that initiates computation of dynamically determined nodes (optimizing frequently used and latency causing nodes).

@Volker Lorrmann does https://github.com/legout/flowerpower have anything here?