Host management policy for energy efficient dynamic allocation

Patricia Stolf (1, 2) , Damien Borgetto (1) , Mael Aubert (1)

1 IRIT-SEPIA - Système d’exploitation, systèmes répartis, de l’intergiciel à l’architecture
2 UT2J - Université Toulouse - Jean Jaurès

Nowadays, reducing the energy consumption of large scale and distributed infrastructures has truly become a challenge for both industry and academia. Dynamic job allocation and resources provisioning is important to fit users requirements. The aim is to minimize the number of hosts utilized in order to reduce the energy consumption while maintaining a correct level of quality of service for the users. Green leverages like migration and on/off actions have cost overheads. Switching on and off a host has a cost: it takes time and it consumes power. These actions have to be optimized and should anticipate load variations. In this paper we investigate host management linked with the allocation and reallocation algorithm in order to optimize the number of powered on hosts and to reduce the overheads. We propose an original host management algorithm based on a genetic algorithm. Our approach has been implemented in DCWoRMS simulator and compared with other heuristics.

Mots clés: Energy consumption; Host management; Green leverages; Scheduling

Domaines: Réseaux et télécommunications [cs.NI]; Système d'exploitation [cs.OS]; Architectures Matérielles [cs.AR]; Systèmes embarqués

Patricia Stolf, Damien Borgetto, Mael Aubert. Host management policy for energy efficient dynamic allocation. 1st International Workshop on Resilience and/or Energy-aware techniques for High-Performance Computing (RE-HPC 2016) in conjonction with 7th International Green and Sustainable Computing Conference, Nov 2016, Hangzhou, China. pp. 1-7. ⟨hal-01692714⟩