Passive network latency monitoring

摘要

A method measures a resident delay for each port in a node in a network and a peer delay between each pair of neighbor nodes in the network. From these resident delays and peer delays, latency between each pair of neighbor nodes in the network is determined. The method includes weighting a route for a data packet going through the nodes in the network using the determined latencies. Each node includes a switch having switchable connections and is configured by a controller to send probe packets from an output port to a port in a neighbor node. The packet may include a time stamp and an identifier.

主权项

1. In a computer network having a plurality of nodes and a plurality of sub-networks, a method comprising:
measuring a first resident delay at each of a selected set of ports at selected nodes in the network; measuring a first peer delay between each pair of the selected nodes that are neighbors to each other; determining a first latency between each of the pairs of the selected nodes that are neighbor to each other; measuring a second resident delay for each sub-network; measuring a second peer delay between each pair of sub-networks; and determining a second latency between each pair of neighbor sub-networks; and weighting a route that a data packet goes through using the first and second latencies.