Streaming Anomaly Detection for Big Data / Internet of Things

The problem of analysis and anomaly detection in time series data arises in many applications, and is especially problematic when systems must cope with high scalability, volume, frequency, and cardinality requirements. We present a system for ingesting and analyzing time series data under such constraints with better performance characteristics than many currently-used systems.

Transcript

1. Streaming Anomaly Detection for Big Data / Internet of Things Adam Drake FOSSAsia 20160319 @aadrake http://aadrake.com
2. @aadrake #hacker or @aadrake #ThoughtLeader
3. Background • Applied mathematics • Startups • Many industries
4. Time Series is the New Big Data https://www.mapr.com/blog/time-series-data-new-big-data
5. https://blogs.technet.microsoft.com/machinelearning/2014/11/05/anomaly-detection-using-machine-learning-to-detect-abnormalities-in-time-series-data/
6. Graphite Prometheus (SoundCloud) Nagios BreakoutDetection (Twitter) Surus (Netflix) Atlas (Netflix) Skyline (Etsy) Bosun (Stack Exchange)
8. https://grey-boundary.io/the-architecture-of-clustering-graphite/
10. metrics kaas
11. DON'T DO THIS
12. https://blogs.technet.microsoft.com/machinelearning/2014/11/05/anomaly-detection-using-machine-learning-to-detect-abnormalities-in-time-series-data/