"Think" If you need higher concurrency

Q: Imagine a scenario. GPs dot, very often. The post is immediately uploaded to the backend. The backend needs to be displayed to the front end immediately.

Two system AB.

A system A that is specifically responsible for depositing to a redis cluster.

A system B, specifically responsible for reading Redis clusters at intervals, and then transferring them to MySQL for a long time. (20s interval, or, tolerate more intervals of loss.) ）

System B is only for long-lasting and lasting.

A system is responsible for "display" and "upload", i.e. "read", "write"

"Write" as a bottleneck.

1. Pyhton+c + redis + nginx + http

2. Python+c + Redis + nginx + socket

3. Openrestry (also appears to have C) + Redis + http

4. Openrestry (also appears to have C) + Redis + tcp\udp

As much as possible to weaken the performance impact of the framework, put the pressure on the Redis cluster (client hash cluster).

Redis operation and maintenance capabilities

Openrestry operation and maintenance capability

"Think" If you need higher concurrency