[LITMUS^RT] [litmus-rt] Experiment scripts

[Björn Brandenburg]

On Apr 12, 2013, at 11:33 PM, Glenn Elliott <gelliott at cs.unc.edu> wrote:

> Jonathan!  You didn't do yourself justice with your explanation!
> 
> Jonathan's tool suite can:
> 1) Automated overhead measurement: (step 1)  Buy a new system.  (step 2) Install Litmus and Jonathan's tools.  (step 3) Run a script with default parameters.  (step 4) A while later, you get overheads.
> 2) Automated Litmus testing.  You give the scripts a few parameters, and it puts the various Litmus schedulers through their paces.
> 3) Framework for repeatable experiments.  If you want to make your experiments repeatable for others, you can give them a few input files and parameters.  The scripts do the rest.
> 
> -Glenn
> 
> On Apr 12, 2013, at 5:26 PM, Jonathan Herman <hermanjl at cs.unc.edu> wrote:
> 
>> Running experiments in LITMUS was hard for me to figure out. I created a repository of experiment scripts for my successors at UNC to make their lives easier. These are based off the scripts used by Chris and I in our paper submissions. I made them as user-friendly as possible, so that I will get as few emails as possible. I've also integrated a few ideas from other people at UNC, included instant messaging (Mac) when an experiment completes, and emails (Glenn) when all experiments complete.
>> 
>> The scripts can be found here:
>> https://github.com/hermanjl/experiment-scripts
>> 
>> I think it would be cool if a repository like this was actively maintained. Anyone want to take mine? Currently, it is very difficult for casual readers to re-create the experiments presented in our papers. It's also too hard to quickly test, debug, and evaluate changes to Litmus, so people just don't do it.


Thanks a lot Jonathan! This looks indeed really great. I would love to adopt these as the "official" LITMUS^RT tracing setup. 

One question regarding the overhead experiments—do the scripts perform any outlier filtering?

Thanks,
Björn