Perpetual Computation on Intermittently Powered Embedded Systems*
(*Batteries Not Included)

Monday April 2, 2012 at 4:15, Steele 006
Kevin Fu
University of Massachusetts Amherst
Sponsored by the CS Colloquium

Abstract

Kevin Fu University of Massachusetts Amherst

Did you hear the one about how many batteries it takes to turn on a Turing machine? None! It's outside the model of computation. Yet it's rather difficult to compute without power. This problem is particularly poignant on tiny embedded systems that resemble smart dust. Batteries overshadow the size of the computer itself. So why not remove the battery?

This talk explores the problems and approaches for computing under such extreme conditions: intermittent power and reboots every few hundred milliseconds. The approaches include low-voltage probabilistic storage on microcontrollers with NOR flash memory (Half-Wits) and energy-aware checkpointing of program state (Mementos). Writing to embedded flash memory at below-specification voltages can reduce energy consumption, but leads to stochastic behavior. The Half-Wits system uses software-only coding algorithms to restore reliability on unmodified hardware by exploiting the electrically cumulative nature of half-written data in write-once bits. Mementos is a software tool that automatically instruments programs with energy-aware checkpoints to protect RAM and registers from power failures. A suite of compile-and run-time tools help to transform long-running programs into interruptible computations. These techniques rethink basic conventions and representations of information for computing platforms that approach the limits of the digital abstraction.

Bio

Kevin Fu is an Associate Professor of Computer Science and adjunct Associate Professor of Electrical & Computer Engineering at the University of Massachusetts Amherst. Prof. Fu aims to make embedded computer systems smarter: better security and safety, reduced energy consumption, faster performance. His contributions on trustworthy medical devices and computational RFIDs appear at ACM ASPLOS, ACM SIGCOMM, ACM TECS, ACM/USENIX HotPower, ACM/USENIX MobiSys, IEEE Security & Privacy, IEEE TC, USENIX FAST, USENIX Security, Financial Cryptography, and the Institute of Medicine. Prof. Fu served as a visiting scientist at the Food & Drug Administration, the Beth Israel Deaconess Medical Center of Harvard Medical School, and MIT CSAIL. He is a member of the NIST Information Security and Privacy Advisory Board. Prof. Fu received a Sloan Research Fellowship, NSF CAREER award, and best paper awards from USENIX Security, IEEE S&P, and ACM SIGCOMM. He was named MIT Technology Review TR35 Innovator of the Year. Prof. Fu received his Ph.D. in EECS from MIT. He also holds a certificate of achievement in artisanal bread making from the French Culinary Institute.

URLs

• http://spqr.cs.umass.edu/moo
• http://spqr.cs.umass.edu/half-wits
• http://spqr.cs.umass.edu/mementos