When I run a lot of actors on every host to achieve massive data transmission, it costs more the six hours on 6000000 cores.
Each actor need to create streams and send to a specified host or actor and all of them compete for real computing resources (single core).
How can I make the most of real computing resources (physical multi-core)?

When I ran several actors in a host with multi-cores, I found that it takes a lot of time, because the simulation almost ran on a physical core. So I want to run the code in parallel.