ACCS 2012 - ACCS Workshop on Adaptive Computing

Adaptability in both organisms and humans has been understood in terms of three qualities: the ability of an organism to survive environmental vagaries, to spread to new habitats, or to respond in intelligent/clever ways to changes in its surroundings. While these observations give a glimpse into organisms’ and human Adaptability, it would be interesting to figure out what this means in the context of machine behavior.
Adaptive Computing is a set of dynamic or run-time techniques to deal with the time-variant objectives of one or more users/applications running on a System or a System of Systems (SoS). A goal in this context may be to maximize the duration of operability before the first detectable fault occurs, or it may be to reconfigure the SoS to change its performance levels. Other possible changes may be architectural or functional or parametric characteristics. These adaptations or reconfigurations may occur any time during the normal operation. Traditional fault tolerance can be considered a sub domain of Adaptive Computing. Discussion at the workshop is expected to bring clarity to the concept of adaptability.
Adaptive Computing encompasses reconfigurable computing (fine/coarse grain). We opine with reasonable evidence, that Adaptive Computing is also a superset of reconfigurable computing. It is about on-the-fly reorganization of either hardware or software or both to achieve targeted performance, thermal and power management, reliability, and availability. Manifestations/Properties include age combative actions, energy efficiency, redundancy exploitation, resilience (to noise, power supply fluctuations, etc), robustness, and varying degrees of reconfigurability. Proper metrics and taxonomies are important and should be discussed at the Workshop. The impact of nanoscale technology on Adaptive Computing should be explored.