There is no evidence that the previously envisioned method is better, and the ACK and SYN processes are more likely to occur, increasing network congestion. It is also fatal that if more requests come in than the LRU cache, the message is discarded and cannot be delivered to other nodes. (as a result of rough testing, I found that messages were not spread evenly and were dropped)
Implement the logic to convert the pull method and store the message to be delivered to other nodes in memory.
There are still problems to be solved. For example, there are issues such as what size the pipe that transmits data to the application should be, and when sending a response to a pull request, if the size is too large, It need to split the packet and send it.
There is no evidence that the previously envisioned method is better, and the ACK and SYN processes are more likely to occur, increasing network congestion. It is also fatal that if more requests come in than the LRU cache, the message is discarded and cannot be delivered to other nodes. (as a result of rough testing, I found that messages were not spread evenly and were dropped) Implement the logic to convert the pull method and store the message to be delivered to other nodes in memory. There are still problems to be solved. For example, there are issues such as what size the pipe that transmits data to the application should be, and when sending a response to a pull request, if the size is too large, It need to split the packet and send it.