Thermal-aware scheduling for data parallel workloads on multi-core processors

Author

Hengxing Tan ; Ranka, Sanjay

Author_Institution

Dept. of Comput. & Inf. Sci. & Eng., Univ. of Florida, Gainesville, FL, USA

fYear

2014

fDate

23-26 June 2014

Firstpage

1

Lastpage

7

Abstract

The main objective of this paper is develop techniques to maximize the throughput of all cores so that both power and temperature constraints met for a multicore processor. The key levers that we have used for this optimization are the clock frequency as well as throttling of cores. The latter shuts off cores during inactive periods. We propose multiple heuristics for maximizing the data parallel throughput on multi-core processors with different frequency scaling schemes.

Keywords

multiprocessing systems; parallel processing; power aware computing; clock frequency; core throttling; data parallel workloads; frequency scaling schemes; multicore processors; power constraints; temperature constraints; thermal-aware scheduling; throughput maximization; Computational modeling; Equations; Load modeling; Mathematical model; Multicore processing; Steady-state; Throughput; Data Parallel; Dynamic Voltage and Frequency Scaling; Load Balancing; Multi-core Processors; Thermal Aware Computing; Workload Distribution;

fLanguage

English

Publisher

ieee

Conference_Titel

Computers and Communication (ISCC), 2014 IEEE Symposium on

Conference_Location

Funchal

Type

conf

DOI

10.1109/ISCC.2014.6912515

Filename

6912515