Reproducing and Solving a Fallback Issue in TCP Prague

This repo contains the experiment code, plotting code and experiment data of the paper Reproducing and Solving a Fallback Issue in TCP Prague.

Reproducing the Experiments

You can reproduce the experiments by running the prague Linux kernel. To run a single instance launch the run.py file in the experiment folder. All command line flags are documented in args.py and the topology is documented in topo.py.

To reproduce a specific experimental setup, use the scripts in the start-scripts subfolder:

• For chapter 3 use reproduce.sh.
• For chapter 4 use extend.sh.
• For chapter 5 first either apply bpftrace-scripts/noinline.diff to the linux kernel, compile it, and run it, or follow the instructions in the three bpftrace files in the bpftrace-scripts folder to change the function being traced. Then use trace.sh.
• For chapter 6 first run the updated prague Linux kernel and then run extend.sh. To run the additional traced run follow the setup instructions for chapter 5 and then run extend-traced.sh.

Using Our Experiment Data

You can retrieve our raw experiment data from the Github release of this repo. The data folder contains a compressed version of our data. To use our plotting notebooks, you don't need to download the raw experiment data. Our raw data was compressed by running the convert.sh script.

Plotting

All plots in the paper were created using the notebooks in the plotting folder.

• For Figure 1 we used explain.ipynb
• For Figure 3 we used reproduce.ipynb
• For Figure 4 and 5 we used extend.ipynb

All plots created are in the images subfolder. You can also find additional plots in that folder which were discussed in the paper. For example plotting/images/extend/patched/classic_ecn.pdf visualizes the bpftrace output of the additional experiment run described in chapter 6.