Abstract

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 extremely difficult to compute without power. Perpetual computing is hard.

This talk will describe recent results on software techniques for low-voltage probabilistic storage on microcontrollers with NOR flash memory (“Half-Wits” at USENIX FAST ) and energy-aware checkpointing on transiently powered, embedded computers (“Mementos” at ACM ASPLOS ). Although embedded systems continue to shrink in size and power consumption, batteries have become larger and heavier than the computer itself. So get rid of the battery. A computational RFID subsists on eight orders of magnitude less energy than a typical AA battery. This lack of energy leads to two research challenges: how to reliably store data in non-volatile memory at low cost and low voltage, and how to cope with the frequent and complete loss of volatile memory on transiently powered computers.

The Half-Wits work analyzes the stochastic behavior of writing to embedded flash memory at voltages lower than recommended by a microcontroller’s specifications to reduce energy consumption. Our software-only coding algorithms enable reliable storage at low voltages on unmodified hardware by exploiting the electrically cumulative nature of half-written data in write-once bits (half-wits). Measurements show that our software approach reduces energy consumption by up to 50% for common workloads.

Mementos automatically instruments programs with energy-aware checkpoints to protect RAM and registers. A suite of compile- and run-time tools help to transform long-running programs into interruptible computations. Contributions include an energy-aware checkpointing system for MSP430 family of microcontrollers, and a trace-driven simulator of transiently powered RFID -scale devices.