Self-aware Execution Environment Model (SAE2) for the Performance Improvement of Multicore Systems

Abstract

Multicore systems are known for operating in a dynamic execution environment. Various conventional approaches/efforts carried out so far towards achieving the dynamism have gradually become obsolete. And there is a dire need to find out and integrate novel paradigms, research theories (such as self-awareness) in Multicore systems to find possible performance optimization alternatives. Self-awareness is one of the important principles of autonomic computing and has been shown a remarkable hope in building dynamic systems. Taking the inspiration from such systems, in this paper a “Self-aware Application Execution Environment” (SAE2) has been proposed. The aim of the SAE2 model is to explore and find out the impact of “self-awareness” in the performance of Multicore systems. The “SAE2 model” is driven through the “autonomic computing principles” and exploits parameter tuning and tradeoff attributes to leverage the Multicore system's potential. The proposed model has a feedback-based mechanism, where an application could interact with the system and signal for various performance issues (at run time) and get the inputs to get adopted as per the system resources availability. A novel “application cooperative behaviour” has been introduced to address various performance issues of Multicore systems.