Download report
Eric Prokop, Peiqian Li
Original paper: Ben Zhang, Xin Jin, Sylvia Ratnasamy, John Wawrzynek, and Edward A. Lee. 2018. AWStream: adaptive wide-area streaming analytics. In Proceedings of the 2018 Conference of the ACM Special Interest Group on Data Communication (SIGCOMM ’18). ACM, New York, NY, USA, 236-252. DOI: https://doi.org/10.1145/3230543.3230554
AWStream (Adaptive Wide-Area Streaming Analytics) [7] is a framework for streaming data analytics applications to smartly trade off between latency and application performance in the face of limited network bandwidth. It aims to simplify application development by providing an interface through which developers can specify configuration “knobs” that control how data is degraded when bandwidth is constrained. AWStream learns the best performing configuration for a given available bandwidth, and uses that to adjust the data send rate when it detects congestion in the network.
We present our efforts to reproduce Figure 12(a) from the original AWStream paper, which compares latency and accuracy for a pedestrian detection application both running with and without AWStream. We also show our reproduction of two intermediate results, including an offline-trained profile and a time series plot for throughput, latency and accuracy.
Reproducing Network Research 