We are pleased to announce the recipient of the ICDE Influential Paper Award:
Predicting Multiple Metrics for Queries: Better Decisions Enabled by Machine Learning
Archana Ganapathi, Harumi A. Kuno, Umeshwar Dayal, Janet L. Wiener, Armando Fox, Michael I. Jordan, David A. Patterson
Today it is taken for granted that machine learning techniques are used for classification, prediction, optimization etc. in a wide variety of domains. 10 years ago, shortly after the UC Berkeley RAD Lab was formed, it was much less obvious that this paradigm was on the horizon. At the time, the use of distributed/parallel data management systems for large-scale (business) analytics was an emerging and challenging use case, making it a compelling problem to solve for in 2009.
We believe the collaboration between RAD Lab and HP Labs was among the first to systematically investigate how machine learning could be applied to operational problems in large software installations – both systems software in datacenters and complex application stacks such as databases. The data engineering methodology presented in our paper has been a foundational framework for timely and actionable decision making across systems and problem domains in industry. In fact, a large number of companies, including Splunk, Anodot, BigPanda, Mixpanel, Avora and Amplitude to name a few examples, have emerged and thrived by following the path we set – using machine learning to understand system behavior.