In my case, the working loads have different sources and sinks than the tested traffic, so I need to do it in the experimentation framework itself (bunch of python scripts). And I get a lot more flexibility in term configuring those parallel iperf sessions (or MGEN). So, personally, I'm not seeing myself ever using the workload feature in this project.
I think the workload feature is great for simple usecases, a good question for you is how much complexity you put into it and what's the limit. For example, what about on-off periodic workload ? Or Poisson traffic ?
Last edit: Jean Tourrilhes 2023-05-12
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
There is the flows code too. Have you tried it? It's based on python's asyncio and ssh.
I've thought about weighted graphs for both writes and delays for the working load. That probably would be the way to realize different patterns. There is the periodic-burst option which probably could be generalzed too.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
In my case, the working loads have different sources and sinks than the tested traffic, so I need to do it in the experimentation framework itself (bunch of python scripts). And I get a lot more flexibility in term configuring those parallel iperf sessions (or MGEN). So, personally, I'm not seeing myself ever using the workload feature in this project.
I think the workload feature is great for simple usecases, a good question for you is how much complexity you put into it and what's the limit. For example, what about on-off periodic workload ? Or Poisson traffic ?
Last edit: Jean Tourrilhes 2023-05-12
There is the flows code too. Have you tried it? It's based on python's asyncio and ssh.
I've thought about weighted graphs for both writes and delays for the working load. That probably would be the way to realize different patterns. There is the periodic-burst option which probably could be generalzed too.