limit the frequency of findnode messages

[zeusoo001]

Background

Currently, the neighbor discovery algorithm runs once every 7.2 seconds (KademliaOptions.DISCOVER_CYCLE), and each time it sends a findnode message to 24 different nodes. So theoretically, a findnode message can be sent to the same node at most once within 7.2s. However, there is no frequency limit when processing findnode messages. The code is as follows:

public void handleFindNode(FindNodeMessage msg) {
  List<Node> closest = kadService.getTable().getClosestNodes(msg.getTargetId());
  sendNeighbours(closest, msg.getTimestamp());
}

To avoid unnecessary network bandwidth consumption, the frequency of findnode messages needs to be limited.

Rationale

To avoid network bandwidth consumption caused by high-frequency neighbor requests, it is necessary to limit the frequency of neighbor requests. The minimum time request interval for each channel can be set to KademliaOptions.DISCOVER_CYCLE / 2. The findnode message is replied to at most once within the minimum time interval.

[jakamobiii]

It is indeed necessary to do a strict check of the message. How to implement this function?

[zeusoo001]

@jakamobiii Add a field to record the time when the findnode message is received.

private long rcvFindNodeMsgTime;

If the next request time is less than the threshold (which can be set to KademliaOptions.DISCOVER_CYCLE / 2), no neighbors message will be responded to.

1. Check whether the time of receiving the findnode message exceeds the threshold, and return directly if it exceeds.
2. Update the value of rcvFindNodeMsgTime.

public void handleFindNode(FindNodeMessage msg) {
  long now = System.currentTimeMillis();
  if (now - rcvFindNodeMsgTime < threshold) {
    return;
  }
  rcvFindNodeMsgTime = now;
  List<Node> closest = kadService.getTable().getClosestNodes(msg.getTargetId());
  sendNeighbours(closest, msg.getTimestamp());
}

[jakamobiii]

Good, which version will support this feature?

[zeusoo001]

@jakamobiii The next version v2.2.6 will support this feature.