David Irwin
Research Scientist
Department of Computer Science
Office: 220
Email: irwin [at] cs [dot] umass [dot] edu
Phone: 413-545-4463
Research
Modern society now relies on complex cyber-physical systems (CPSs) that coordinate physical, computational, and network elements to perform important societal functions, such as reducing energy consumption, limiting carbon emissions, monitoring strategic infrastructure, detecting severe weather, optimizing traffic patterns, etc. As a result, understanding how to design CPSs to be robust, efficient, secure, and manageable is critical. The basic structure of future CPSs is beginning to emerge: real-world physical elements, including cars, buildings, phones, etc., continuously sense and transmit operational and environmental data via the Internet to remote software systems hosted at various data centers ``in the cloud"; these systems in-turn process the data to inform or directly control real-world physical processes. Multiple technology trends are converging to shape this vision of CPS. Most importantly, the amount of available real-world sensor data is growing exponentially, requiring substantial computational resources for processing and storage. In parallel, cloud data centers are driving down the cost of computation and storage by leveraging statistical multiplexing and massive economies of scale. Finally, the network connectivity linking the real world with the cloud is becoming more pervasive, plentiful, and flexible with the emergence of 3G/4G wireless connectivity, high-bandwidth broadband, and programmable networks, e.g., via GENI, OpenFlow, etc. Thus, CPSs are rapidly evolving from multiple isolated end-systems to networks of diverse physical elements communicating with remote clouds to coordinate their collective actions.
Since CPSs encompass many different types of systems, a key research challenge for CPS software systems is defining a common platform for designing and implementing novel abstractions, techniques, and algorithms for a diverse set of CPS environments. Just as software systems for smart phones have enabled developers to build thousands of innovative data-driven ``apps," the goal of future CPS software systems should be to enable developers to build diverse apps using real-world sensors and actuators, such as a building-based app to lower the electric bill or a car-based app to improve fuel-efficiency. As with mobile apps, achieving this goal would open up CPSs to the type of widespread disruptive innovation that rapidly accelerates technology advancement. My research centers on realizing this vision by studying the design, implementation, and evaluation of next-generation cloud-based CPSs. My prior work explores the design of systems for the wide range of environments utilized by CPSs, including embedded sensor systems, cloud data centers, and Internet-scale networked systems, while also applying systems techniques to design novel CPSs, such as energy-optimized smart buildings and virtualized camera networks. Despite the different settings, my research methodology follows a consistent theme: elucidating fundamental systems design principles and techniques by building, deploying, and evaluating real hardware/software prototypes.
My research encompasses CPSs as well as topics that will influence future CPSs, including Green Computing, Embedded Sensor Systems, and Next-generation Clouds and Networks.
Cyber-Physical Systems leverage computation and networking resources to sense, actuate, and adapt real-world physical systems. My recent research focuses on using computation and sensing to reduce society's energy footprint, primarily by optimizing the energy consumption of smart electric grids and buildings.
Green Computing explores methods for reducing the energy footprint of computation. My research focuses on reducing the energy footprint and carbon emissions of the massive data centers now common at large institutions. In particular, recent research examines how to integrate intermittent power sources into data centers.
Embedded Sensor Systems are the foundation of CPSs, since they sense and transmit actionable data about the physical world in real time. My research focuses on designing perpetual sensor systems that are continuously operational, maintainable, and retaskable despite unknown energy constraints, failure rates, and application scenarios.
Next-generation Clouds and Networks must accommodate the I/O-intensive ``big data" requirements of future CPSs. My research focuses on system architectures and techniques for improving resource management in virtualized cloud data centers and next-generation, programmable networks, such as GENI.
See a recent research statement (updated December 2011) or check out my publications for more information.