Calculating performance metrics using trace.json for simulation in UnetStack3

Question

I am working on the tool to calculate different performance metrics (like average end-to-end delay, throughput, packet delivery ratio, etc.) for the simulation of underwater networks in UnetStack3. I have done an implementation in python that parses the trace.json and calculates end-to-end delay. However, it works only for topology with one-hop communication, as I have considered the MessageID of the events. Further, I analyzed the implementation of the VizTrace tool in Julia and tried to extend the implementation. However, I am unable to figure out how to co-relate events that occur in different nodes for calculating performance measures in a multi-hop topology. Please let me know what approach I should follow with Python and with the vizTrace.

Answer 1

Every event entry in the trace.json file contains a few useful pieces of information to help you associate events across nodes:

• component: name and class of the agent, along with the node on which it is running
• threadID: unique identifier within a node that associates related events together.
• stimulus: contains messageID of the message that caused this event.
• response: contains messageID of the message that was sent in response to this event.

For more details, see https://blog.unetstack.net/whats-new-in-UnetStack-3.3

Tracing an event through the agents in the node simply involves collating the events with the same threadID. In order to trace an event across nodes, you need to look at the messageID of the response messages, and find the equivalent stimulus message (same messageID) on the next node. Then you do the same from that node to the following one, until you reach the destination.

If you are using the HalfDuplexModem simulation model, then these messages that go across nodes (and hence across threadIDs) will be the HalfDuplexModem$TX messages. Example: https://blog.unetstack.net/assets/img/mermaid-diagram-20210408195013.svg