Zoom: https://illinois.zoom.us/j/87420079461?pwd=Q2VBOTN6WkZVdUJTTVdNcjFZMW1pQT09
Abstract:
Networked systems are becoming increasingly large and complex. These critical infrastructures have inevitable failures that make their management and reliability challenging. System admins often have to troubleshoot such infrastructure issues using commonly available telemetry and have to manually sift through lots of data in the hope of finding a proverbial needle in a haystack. I have developed systems to automatically infer the root cause of such issues, using ML and algorithmic techniques, addressing challenges of large scale and complexity of real-world infrastructure. I will present three such systems that tackle typical problems that systems admins have to diagnose- e.g. why the CPU usage of a VM is currently high, why some packets are getting dropped, or why a client is experiencing high response times. A unifying theme across these systems is that they employ powerful reasoning techniques to derive insights from available monitoring information.
Bio:
Vipul Harsh is a Ph.D. candidate with Brighten Godfrey at the University of Illinois at Urbana-Champaign He is broadly interested in networked and distributed systems. His work spans failure diagnosis in networked systems, datacenter topology, distributed monitoring, and parallel algorithms. His work has been published in top-tier conferences in Computer systems and networks, and one of his projects has been adopted into a product at VMware.
https://wiki.illinois.edu/wiki/display/grad/PILOT+SEMINAR
