This is the artifact repo for the paper Caravan: Practical Online Learning of In-Network ML Models with Labeling Agents (to appear at OSDI 2024).
The paper artifact includes Caravan's software emulation and the FPGA+P4 testbed. The emulation folder contains the source code to automatically run key experiments from the paper and reproduce the corresponding figures (i.e., Figures 4a, 4b, 5--9, 10a, 10c). The testbed folder contains the new code changes and the instructions to set up and run the FPGA+P4-based evaluations for Caravan.
Note for AE reviewers: The
testbedfolder is not required during artifact evaluation since it requires specialized hardware that is difficult for us to accommodate; however, anyone with the necessary hardware can try out the instructions.