Reminds of the way LMAX achieve high throughput and low latency on their trading exchange. One of the key points is that they use parallelism but not concurrency. Keep your threads busy doing one thing only and avoid synchronising with other threads. No blocking, no queues and they achieve >100K messages/s.
Where they do need pass messages between threads they use the famous LMAX Disruptor which keeps data copying and blocking to a minimum.
Where they do need pass messages between threads they use the famous LMAX Disruptor which keeps data copying and blocking to a minimum.