ABSTRACT
Along with the surging service demands in the cloud, power provision infrastructure of Internet Data Centers (IDCs) has brought dramatically increasing capital cost. To enlarge the size of IDCs with lowest cost, power management of computing facilities has attracted many attentions in recent. A large portion of applications running on data centers are data-intensive and throughput-preferredMapReduce is one of them enjoying widely deployment. However the critical power peak problem in MapReduce clusters, which actually limits the cluster's size, has been overlooked. We study the power peak problem in MapReduce system and investigate the reason causing it. We design an adaptive approach to regulate power peaks. Evaluation result shows that our proposed method can effectively smooth the power consumption curve by reducing the peak value for 20% with little overhead in performance, and in turn extending the maximum size of the cluster with 25% under the same power budget.
- Y. Chen, S. Alspaugh, D. Borthakur, and R. H. Katz. Energy efficiency for large-scale mapreduce workloads with significant interactive analysis. EuroSys, 2012. Google ScholarDigital Library
- J. Dean and S. Ghemawat. Mapreduce: Simplified data processing on large clusters. Operating Systems Design and Implementation (OSDI) 04', pages 137--150, 2004.Google Scholar
- X. Fan, W. dietrich Weber, and L. A. Barroso. Power provisioning for a warehouse-sized computer. In Proceedings of the ACM IEEE International Symposium on Computer Architecture (ISCA), 2007. Google ScholarDigital Library
- Hadoop. http://hadoop.apache.org/.Google Scholar
- J. Hamilton. Internet-scale service infrastructure efficiency. Keynote at the International Symposium on Computer Architecture (ISCA), 2009. Google ScholarDigital Library
- T. Horvath and K. Skadron. Multi-mode energy management for multi-tier server clusters. The Seventeenth International Conference on Parallel Architectures and Compilation Techniques (PACT), 2008. Google ScholarDigital Library
- J. Leverich and C. Kozyrakis. On the energy (in)efficiency of hadoop clusters. Workshop on Power-Aware Computing and Systems (HotPower), 2009.Google Scholar
- D. Meisner, C. M. Sadler, L. A. Barroso, W.-D. Weber, and T. F. Wenisch. Power management of online data-intensive services. In Proceedings of the ACM IEEE International Symposium, on Computer Architecture (ISCA), 2011. Google ScholarDigital Library
- N. Zhu, L. Rao, X. Liu, J. Liu, and H. Guan. Taming power peaks in mapreduce cluster. Proceedings of the ACM SIGCOMM 2011 conference on SIGCOMM, 2011. Google ScholarDigital Library
Recommendations
Handling more data with less cost: taming power peaks in mapreduce clusters
APSys '12: Proceedings of the Third ACM SIGOPS Asia-Pacific conference on SystemsAlong with the surging service demands in the cloud, power provision infrastructure of Internet Data Centers (IDCs) has brought dramatically increasing capital cost. To enlarge the size of IDCs with lowest cost, power management of computing facilities ...
More is Less, Less is More: Molecular-Scale Photonic NoC Power Topologies
ASPLOS'15Molecular-scale Network-on-Chip (mNoC) crossbars use quantum dot LEDs as an on-chip light source, and chromophores to provide optical signal filtering for receivers. An mNoC reduces power consumption or enables scaling to larger crossbars for a reduced ...
More is Less, Less is More: Molecular-Scale Photonic NoC Power Topologies
ASPLOS '15: Proceedings of the Twentieth International Conference on Architectural Support for Programming Languages and Operating SystemsMolecular-scale Network-on-Chip (mNoC) crossbars use quantum dot LEDs as an on-chip light source, and chromophores to provide optical signal filtering for receivers. An mNoC reduces power consumption or enables scaling to larger crossbars for a reduced ...
Comments