Closing this - will write a new impl.

I think there's a fundamental problem with both this solution here and the existing beacon processor - Both async and blocking tasks are handled the same way e.g., some async tasks that wait on network I/O may hold a worker slot unnecessarily.

The async task issue is interesting, maybe we should allow async tasks to go unbounded, and let Tokio's thread pool sort them out?

Alternatively we would decrement the number of active workers every time we yield, but I don't think this sounds easy to do

Yeah i think we could potentially try that.

Should also keep in mind @paulhauner's comment below - there are some expensive async tasks (e..g. reconstruction - although we might be able to split this one into two tasks - blocking reconstruction + async block import)

I think there's two types of async tasks:

Async tasks with very little computational load (e.g. waiting for a response from a server)
Async tasks which have high computational load (e.g. those that might spawn a few blocking threads and await them)
I think (1) doesn't need to take a worker slot, however I think it makes sense that (2) takes a slot since we don't want too many of those running simultaneously.

Yeah perhaps we could make the blocking tasks started by async tasks go through the beacon processor so that they are tracked and bounded

Yes, so to summarise:

BeaconProcessor

Async task

• Keep computation light
• Any blocking computation go through the beacon processor
• Not bounded by max workers (num_cpus) but probably still need a queuing system to make sure it doesn't go unbounded and cause memory issues

Blocking task

• Gets allocated 1 thread by default
• For heavy tasks that requires rayon, allowWorkTypes to acquire more than 1 worker (N) - when executing, create a scoped rayon pool and run the parallel tasks within the scope, and release the workers after it completes

CPU allocation

• tokio runtime: num_cpus thread (default)
• BeaconProcessor: running inside the tokio runtime, and maintains blocking workers max_blocking_workers = num_cpus
• rayon: scoped per task

Memory

• Both sync and async tasks are bounded by queues of each individual work type.