Closed carlhoerberg closed 2 years ago
LavinMQ using this branch: Average publish rate: 707774.2 msgs/s Average consume rate: 699375.7 msgs/s
master branch: Average publish rate: 637212.4 msgs/s Average consume rate: 636811.7 msgs/s
10% throughput increase
When only publishing: Average publish rate: 1476317.1 msgs/s
vs. master branch: Average publish rate: 1268211.3 msgs/s
16% increase
Int.from_io will (stack) allocate a buffer and then fill it and then parse, instead now we read as much as possible and then decode each part of a longer byte array.