Usability Analysis of the Faban Framwork

[Stefan Geißler]

Hi everyone,

I am currently evaluating if Faban is usable (and sensible to use) for our scenario and I thought i might eventually save a lot of time by just asking someone who has already worked with the framework and may be able to answer the question. So here we go.

In our scenario, we are going to benchmark applications that do not handle requests but raw network packets. To do so, we will have to use a high performance traffic generator like DPDK-PktGen to generate and send the traffic workload to the system under test. In addition to that, we will need system calibration in order to assess the maximum steady-state load level of the SUT since we need to run various load levels between 0% and 100%.

The question now is, wether we can (or should) use the Faban framework in this type of scenario. As far as I can tell, all benchmarks designed in the Faban context generate load on a request basis (e.g. a http request). In order to realize the incorporation of the external traffic generator, will I have to change or add code to the Faban framework itself or is it possible to implement all needed functionality into the benchmark driver (the plugin .jar file)?

Has anybody already experience with this type of scenario and can maybe provide me with some information on how this is possible with Faban within a sensible amount of implementation work or wether I would have to change loads of code to achieve such thing.

Thank you already in advance!

Regards,

S