batching / pipelining requests for small byte size data

Open kolinfluence opened this issue 2 years ago • 1 comments

smaller sizes bandwidth throughput is much lower. was wondering if can provide a "batching" / pipelining requests through option of time (10ms) or slice size (e.g. 100) items and with size definition for the pipeline (e.g. <1kb) or similar in nature for shmipc requests only. not the fallback uds ones.

certain ops will benefit from the smaller sized transactions.

Jun 21 '23 11:06 kolinfluence

Excellent suggestion. This would be very useful in offline scenarios, and we are considering supporting it in the future. In fact, it is already in our roadmap: https://github.com/cloudwego/shmipc-go/blob/main/ROADMAP.md

Jun 23 '23 12:06 goodbye-babyer