Machine Learning for Log Monitoring

At the November Test in Production Meetup in San Francisco, Larry Lancaster, Founder and CTO at Zebrium, discussed using machine learning to organize and detect patterns in unstructured log data. Unsupervised machine learning models can help teams find the root cause of incidents in their stack and use such insights to prevent future errors. Machine learning can aid in faster recovery from incidents and even avoid them entirely. The talk highlighted the challenges faced when dealing with unstructured log data, including format changes and the need for manual interpretation. Lancaster also shared his vision for a future where smart metrics companies integrate various types of data to perform real anomaly detection.