High-quality data repositories and benchmarks have enabled research on computer vision and language understanding to progress at a rapid clip. However, many ML and advanced analytics applications in the industry use time-series data – not [just] image, video, and text datasets. As such, the gap between industry adoption of ML and academic research continues to widen. To address this problem, Jacob et al. released Exathlon, a comprehensive public benchmark for explainable anomaly detection on high-dimensional time series data. Specifically, Exathlon represents a challenge that many companies face – Spark application monitoring. The benchmark includes a dataset based on real data traces collected from repeated executions of distributed streaming jobs on a Spark cluster (including some jobs that were intentionally disturbed with chaos engineering approaches). It also includes a new benchmarking methodology for anomaly detection and explanation discovery tasks and an end-to-end pipeline to facilitate experimentation.