Software agents
The remote processing framework for portable computer power saving
Proceedings of the 1999 ACM symposium on Applied computing
A discrete-time battery model for high-level power estimation
DATE '00 Proceedings of the conference on Design, automation and test in Europe
ACM Computing Surveys (CSUR)
Fast, flexible, cycle-accurate energy estimation
ISLPED '01 Proceedings of the 2001 international symposium on Low power electronics and design
Operating system support for mobile agents
HOTOS '95 Proceedings of the Fifth Workshop on Hot Topics in Operating Systems (HotOS-V)
Modeling, Analysis, and Self-Management of Electronic Textiles
IEEE Transactions on Computers
| Hi-index | 0.00 |
As technology advances, devices become smaller and cheaper, making it possible to build systems containing large numbers (possibly hundreds, or more) of miniature processing elements. Such platforms, although superficially similar to traditional distributed systems, pose additional unique challenges. Due to the desire to minimize costs, coupled with the sheer numbers of devices, it will be difficult to perform manufacture-time testing, and runtime-failures will likewise be common. One challenge is to efficiently harness the capabilities of large numbers of low power (and relatively low performance) processing elements, in the presence of failures such as depleted battery resources, as well as those due to unpredictable sources (e.g., electrical and mechanical failures). It will however be possible to employ a fraction of the multitude of resources as redundant or spare devices, re-mapping applications onto them, from failing ones.This paper investigates the use of code migration as a general means of performing such application re-mapping, in the presence of intermittent communication and device failures, as well as limited battery resources. A new technique, Pre-Copying with Remote Execution (PCRE), an extension of code migration which enables more efficient application re-mapping in the presence of energy and communication constraints for symmetric applications, is presented.It is shown that PCRE provides a 28.6% improvement in system lifetime and 9.8% improvement in energy efficiency for the applications investigated, over the baseline code migration strategy. Naturally, the re-mapping of applications involves overheads in computation and communication, and PCRE reduces these overheads to within 10% of the ideal case of doubled energy resources.