Question

I use boost.asio async read data, if handle_read takes a lot of time (such as sleep), does it affect other connections?

Answer 1

If you execute io_service::run in one thread, the completion handlers of all the i/o objects associated with that io_service are invoked sequentially. So, if one of them takes too much time, all the queue gets stuck.

To prevent this situation, you can either re-design this completion handler, or associate the i/o object having problematic handlers with its own dedicated io_service running in a separate thread.

(Note that running a single io_serivce in multiple threads wouldn’t solve this problem, as you can’t know how the handlers are distributed among the threads.)