Project Description

Oracle recently gifted our group the world’s largest raspberry pi cluster – which we call Godzilla Pi.   It provides our group with the next generation IoT/Cloud systems for pursuing low-power parallel/distributed computing at scale.  Managing and understanding the behavior of so many devices as a cluster (1025 pi3 devices) poses many systems research challenges. With this project, the ERSP team will design and develop a visual monitoring system that uses AI to identify the individual devices viewed via a camera/ipad and that provides visual clues about device activity and behavior, i.e. which devices are up/down, which devices are experiencing faults/errors, what each device is doing (sleeping, computing, communicating, etc.), and how much heat the system is generating.

The visual display tool that the ERSP team will develop will be made available to users of the cluster via a real time web service and as an intelligent, physical LED light strip that summarizes device-aware cluster status. The choice of colors and graphics will be the design choice of the team and allow for quick understanding of which devices need attention and which ones are in use (to avoid conflicts).  The project will extract information from logs and other system-level operations to inform this display. The monitoring/extraction should not impose significant performance/energy overhead on the device. Stretch goals include measuring and estimating energy consumption  and developing algorithms that use the data to predict when devices are likely to become faulty so that proactive maintenance can be pursued.

The team will learn operating and programming systems tools for monitoring support, analytics and data processing tools for extracting actionable insights from logs, and cluster/cloud management techniques used today in industry and as part of cutting edge systems research in cloud and IoT. They will also learn how to efficiently profile systems at scale and communicate the information to users and administrators of such systems.  Finally, they will use modern software engineering practices. Finally, they will use modern software engineering practices in the team-based development of the project.

 

Team Members

  • Emily Zheng
  • Karen Yuan
  • Ria Singh
  • Shruthi Santhosh Unnithan

 

Professor and Mentors

  • Prof. Chandra Krintz and Prof. Rich Wolski
  • Grad mentors: Animesh Dangwal