Tommy Fan, Austin Poore
Original Paper: Ben Basat, Ran, et al. “Constant time updates in hierarchical heavy hitters.” Proceedings of the Conference of the ACM Special Interest Group on Data Communication. ACM, 2017.
We present and explain our efforts to reproduce key results from “Constant Time Updates in Hierarchical Heavy-Hitters” for CS 244, a graduate networking course at Stanford University. In particular, we run several experiments using code from the original authors to attempt to reproduce their performance and error metrics on our own servers. In addition, we present an evaluation of our own implementation of the algorithm in the one-dimensional setting.
Our work falls into two broad sections: in the first, we use the original authors’ code to run their experiments on our own machine and generate fresh data, which we then use to reproduce four of the figures from the paper detailing error metrics and performance. Then, in the second, we present the empirical results from our own implementation of the one-dimensional algorithm, which were comparable to the results from the authors’ implementation. As a result, we feel very confident in endorsing the paper as being reproducible. As part of our reproduction efforts, we’ve also written some scripts to automate the processes of running all of the experiments and generating the resulting plots which, when packaged with the authors’ original code, will make it even easier for future reproduction efforts to confirm the paper’s findings.
Reproducing Network Research 