Generally speaking, we can calculate the packet delivery time using this formula, Wikipedia:
Latency = Transmission delay + Propagation delay + queuing delay + processing delay
My question is: Is it acceptable to use the value of RTT to be the estimated latency? In fact, the simulation environment I use provides the calculation of RTT so I wonder if it is close enough to use this value to be the estimated latency of a path from source and destination.
Kindly, I'd like to refine this reasoning if it is not correct.
Best Answer
Network latency (what is on topic here) will vary from measurement to measurement. The queuing delay is constantly changing, based on the traffic at the time. With relatively little traffic, you will see pretty good latency, but as soon as someone start downloading a software update or watching a video, your latency can greatly increase until that stops. There are tools, like IP SLA, to constantly measure latency of various protocols, which will be different for different network protocols.
RTT will be more than double the network latency because it includes both directions and the host processing times.
Properly configured and managed QoS will give you the best throughput. Large queues under heavy load will give you the worst latency, so smaller queues with things like RED that will purposely drop packets in order to keep queues from filling and TCP from getting globally synchronized (a bad thing) so throughput can be greatly affected, are used.