The Community for Technology Leaders
2016 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW) (2016)
Chicago, IL, USA
May 23, 2016 to May 27, 2016
ISBN: 978-1-5090-3683-7
pp: 7-16
ABSTRACT
Efficient heterogeneous resource management is one of the many challenges Cloud systems exhibit, because it can lead to increasing the profit obtained by Cloud providers by serving more client requests while fulfilling their requirements with the minimum set of resources. In this context, heterogeneous architectures composed by accelerators, and Field Programmable Gate Arrays (FPGAs) in particular, can enhance the fulfillment of client requirements with less power consumption, as long as they are managed in a smart way. Thus, the HEterogeneous Cloud COmputing (HECCO) framework developed in this work enables the integration and efficient management of FPGAs as co-processing resources within a Cloud Computing paradigm. This framework uses allocation and scheduling algorithms based on classification and prediction models in order to choose the most suitable combination of resources according to the applications and SLA parameters. A proof-of-concept implementation of the HECCO framework over an image processing service case-study is evaluated on a small testbed. Results show how the smart use of FPGAs integrated with conventional computational resources leads to a higher percentage of clients serviced with their QoS fulfilled and even supporting SLAs with more stringent deadlines. Additionally, energy savings exist, which can contribute to reduce the energy footprint of data-centers.
INDEX TERMS
Cloud computing, Field programmable gate arrays, Quality of service, Computer architecture, Resource management, Companies, Proposals
CITATION

J. Proano, C. Carrion and M. B. Caminero, "Towards a Green, QoS-Enabled Heterogeneous Cloud Infrastructure," 2016 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), Chicago, IL, USA, 2016, pp. 7-16.
doi:10.1109/IPDPSW.2016.12
93 ms
(Ver 3.3 (11022016))