Dependent Variables - clarification on csv files

[gentryae]

Data processor:

• does “experiment designers” refer to the people running the tests or the actual experiment designers?
• Is the first .csv “processing latency” the roundtrip time for the message to go from pub to consumer and back? If not, wouldn’t we have. processing latency for both the pub and the sub? Which would we care about (if not both). In other words (just to be suppppperrrr clear) Do we want End to End latency for pub -> sub AND end to end latency for sub -> pub response?
• wouldn’t the second csv file also be two separate files? One for the number of messages sent per second by the pub. another for the number of messages received by the sub?

[jmbeach]

• Experiment designers is me and Layne, essentially. Not the the people running the tests
  • Working to get that to everyone quickly
• Correct. processing latency is the roundtrip time from pub to consumer and back.
  • We're only wanting the total end-to-end latency (pub to sub and back) and are not measuring the individual latencies: from pub to sub and then from sub to pub
  • Originally, I wanted to collect more granular information, but the group (experiment designers) decided it wouldn't be that valuable because at that point, we'd be testing how fast communication is. It's an interesting chat. Here's a snippet:

    I honestly think that should be a WHOLE different experiment, that I’m not sure we have time for. And unless any of these techs are just terrible, your internet bandwidth should be the limiting factor on message speed anyway, right? In a real world scenario

You can read the whole thing here.

~* You're right! I will update the output to reflect that. However, I'll only make it output a separate column for receive throughput instead of an entirely new csv.~

• Actually, correct me if I'm wrong, but since we are only collecting output from the producer side, there will never be a difference in those numbers.
  • For the producer to move on to the next message, it has to get the ack back first. If we were collecting data on both sides, then that would be the case. I'll still update the readme to be more clear.

Thanks for the attention to detail on these questions.

[jmbeach]

Just merged #45