FlyDB optimization problem

[sjcsjc123]

In order to improve the throughput of FlyDB, we have the following options:

1. WAL logging operation, using memory read/write key-value and asynchronous drop disk

2. Introduce message queues, hand over disk drop operations to kafka or mq, and the main process only performs memory reads and writes

3. Enable multiple processes in the background, pass in corresponding processes based on the key, execute data drop operations in the processes, and the main process only performs memory read and write operations

4. If a certain monitoring mechanism is adopted and the asynchronous disk drop is unsuccessful, can a notification be sent to the client before the machine goes offline.

All of the above will have Big data loss status when the machine is offline instantly. If Kafka is deployed in other nodes, the loss will be smaller, but the operation is relatively more complex.

[qishenonly]

good issue！

[qishenonly]

In addition, we are talking about improving throughput here in the case of single machine optimization. We need to optimize single-threaded read and write so that it has better read and write performance.