Assign io_context for each thread. #866
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
The original design was calling
ioc.run()
from multple threads.It worked well but scale out limit is up to about 8 threads.
The pps is 120,000 publish / second (QoS0). It was pretty nice but only
8 cores ware used.
I tested it on AWS 48 vcpu environment.
When I used 48 threads, the result was the same as 8 threads.
I suspected OS or VM limit. But when I ran the two broker processes that
listens different port, then each broker resulted 120,000 pps.
So total is 240,000 pps.
It indicates the limit is not caused by OS or VM.
Finally, I found the bottle neck. It is io_context.
So I updated as follows:
Separated io_contexts.
For timer, for accepting connections, and for MQTT communication.
For MQTT communication, I prepared io_context as the same number of
the threads.
Added server constructor that accepts io_context_getter.