An overview and classification of thermal-aware scheduling techniques for multi-core processing systems

Thermal-aware scheduling of parallel jobs is becoming an increasingly critical issue in software design for computing platforms ranging from embedded systems to large servers and entire data centers. In addition to the obvious implications of unrestricted temperature, such as thermal emergencies resulting in hardware failure, effective thermal management reduces the cost of thermal packaging for processors while achieving long-term savings from reduced secondary cooling costs. In embedded systems, hardware temperature control mechanisms in conjunction with software and other dynamic thermal management (DTM) techniques have been effectively utilized to avoid overheating. Server systems can also optimally take advantage of advanced DTM and efficient task scheduling to exert a positive impact on overall cooling costs as well as on prolonging the life of system components. The aim of this paper is to provide a comprehensive survey of recent research on thermal-aware task scheduling techniques for multi-core systems. The paper starts with the motivation behind managing temperature through task mapping and scheduling and then addresses the problems faced in such an approach, while providing commonly used thermal and task models as well as contemporary temperature control mechanisms of modern multi-core processors. A number of objectives of optimization with different combinations of temperature and performance constraints are elaborated. Also described in the context of above mentioned issues is a structured coverage of recently published research work. This leads to grouping of existing techniques in a taxonomical manner as well as to some conclusive observations that can help in identifying possible new research directions.

► Comprehensive survey of recent thermal-aware task scheduling techniques for multi-core systems.
► Structured coverage of recently published research work and grouping of existing techniques in a taxonomical manner.
► Observations that can help in identifying possible new research directions in the area of thermal-aware scheduling.
► Elaborates objective functions and constraints for optimizing performance, energy and temperature in multi-core systems.