Training: June 20–21, 2016
Tutorials: June 21, 2016
Keynotes & Sessions: June 22–23, 2016
Santa Clara, CA

Robust anomaly detection for real user monitoring data

Ritesh Maheshwari (LinkedIn), Yang Yang (LinkedIn)
11:20am–12:00pm Thursday, 06/23/2016
Infrastructure reimagined
Location: Ballroom AB Level: Beginner
Average rating: ****.
(4.43, 7 ratings)

Prerequisite knowledge

Attendees should have a basic understanding of RUM.


For the past year, LinkedIn has been running and iteratively improving Luminol, its anomaly detection system that identifies anomalies in real user monitoring (RUM) data for LinkedIn pages and apps. Ritesh Maheshwari and Yang Yang offer an overview of Luminol, focusing on how to build a low-cost end-to-end system that can leverage any algorithm, and explain lessons learned and best practices that will be useful to any engineering or operations team. LinkedIn will be open sourcing its Python library for anomaly detection and correlation during the talk.

Topics include:

  • Use cases
  • How to avoid an alert black hole
  • Data processing
  • Overview of Luminol
  • Root cause detection
  • Alerting
  • Success stories
Photo of Ritesh Maheshwari

Ritesh Maheshwari


Ritesh Maheshwari is a performance engineer at LinkedIn, working on making LinkedIn fast using his medley of skills in data and performance analysis, network optimizations, and automation. Before LinkedIn, Ritesh was a performance engineer at Akamai, doing something similar. Ritesh holds a PhD in computer science from Stony Brook University, where he first became passionate about performance while working on computer networks. He is also an alumni of IIT Kharagpur.

Photo of Yang Yang

Yang Yang


Yang Yang is a statistician interested in understanding data and digging signals via statistical modeling. Yang’s work involves developing metrics and methods to track the long-term success of web services and detect key drivers of user engagement and verify them rigorously through statistical analysis. Yang is also keen on developing novel statistical methods to solve real data problems. During her PhD at University of Michigan, Yang proposed new methodologies to analyze panel data with incomplete observations by stochastic optimization techniques and MCMC sampling.