A Proficient Thermal Based Prediction Unit in Multi Core System

Abstract

Multi core system has become a popular Execution platform for many Real time systems. The system is continue to shrink down due to thermal hotspots and temperature gradients, which Are significantly impact system parameters like Reliability, performance, cost and leakage power To improve these parameters in Real time systems/ multi core system a Temperature prediction management (TPM) technique is required for mission/safety critical application running on a multi core chip to avoid possible thermal hazards while meeting the application timing constraints. In this paper, we propose an efficient framework for Temperature prediction management (TPM) with time constraint model (Tracker & WDT) in many core systems where the balanced thermal profile can be achieved by proactive task migration among neighboring cores. And based on temperature prediction unit task migration will be performed among neighboring processing elements. Compared with existing proactive task migration technique ,our approach effectively minimize the thermal hotspots, less migration overhead with negligible performance overhead and it increase the processing speed of multi core systems.