feat: add message byte batching #235
Conversation
Adds a new setting `minSendBytes` that is `undefined` by default. If `undefined` all messages sent through multiplexed streams will be serialized and sent over the wire immediately. If set to a number, it will be used as a byte value, and the serialized bytes of all messages sent during the current tick will be buffered up to that value. Once either the buffer lengths hits that value or the next tick begins, all bytes in the buffer will be sent over the wire.
|
There are some gains to be had for small message sizes though they evaporate once messages sizes get above about 10kb. Partially you'd expect this as the buffered message data is yielded as soon as it goes above The graph below was generated using the benchmarks in https://github.com/ipfs-shipyard/js-libp2p-transfer-performance The test sends 105MB of data from one node to another over TCP divided across a number of messages. The x-axis is individual message sizes, the y-axis is how long it takes to send that much data. The legend on the right is the value of
|
## [7.1.0](v7.0.7...v7.1.0) (2022-11-25) ### Features * add message byte batching ([#235](#235)) ([4e2a49d](4e2a49d))
|
🎉 This PR is included in version 7.1.0 🎉 The release is available on: Your semantic-release bot 📦🚀 |
Adds a new setting
minSendBytesthat isundefinedby default.If
undefinedall messages sent through multiplexed streams will be serialized and sent over the wire immediately.If set to a number, it will be used as a byte value, and the serialized bytes of all messages sent during the current tick will be buffered up to that value.
Once either the buffer lengths hits that value or the next tick begins, all bytes in the buffer will be sent over the wire.