Adaptive Preference-Aware Co-Location for Improving Resource Utilization of Power Constrained Datacenters
- 15 September 2020
- journal article
- research article
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Parallel and Distributed Systems
- Vol. 32 (2) , 441-456
Abstract
Large-scale datacenters often host latency-sensitive services that have stringent Quality-of-Service requirement and experience diurnal load pattern. Co-locating them with best-effort applications that have no QoS requirement has been widely used to improve the resource utilization. However, existing co-location techniques tend to result in power overload problem on power constrained servers. To this end, we propose Sturgeon, which proactively manages resources between co-located applications in a power constrained environment, to ensure the QoS of latency-sensitive services while maximizing the throughput of best-effort applications. Our investigation shows that, at a given load, multiple feasible resource configurations exist and meet both QoS requirement and power budget. To find a feasible configuration with maximum throughput, we establish models to predict the performance and power consumption of the co-located applications. Besides, when the datacenter hosts different types of applications to perform co-location, Sturgeon places applications with their preferable candidates to improve the overall throughput. The experimental results show that at server level Sturgeon improves the throughput of the best-effort application by 25.43% compared to the state-of-the-art technique, while guaranteeing the 95%-ile latency within the QoS target; at cluster level, Sturgeon improves the overall throughput of best-effort applications by 13.74% compared to the baseline.Keywords
Funding Information
- National R&D Program of China (2018YFB1004800)
- National Natural Science Foundation of China (61632017, 61772480, 61872240, 61832006, 61702328)
This publication has 42 references indexed in Scilit:
- Elastic Power-Aware Resource Provisioning of Heterogeneous Workloads in Self-Sustainable DatacentersIEEE Transactions on Computers, 2015
- Using platform level telemetry to reduce power consumption in a datacenter2015
- Bubble-fluxACM SIGARCH Computer Architecture News, 2013
- High-dimensional kNN joins with incremental updatesGeoInformatica, 2009
- The Datacenter as a Computer: An Introduction to the Design of Warehouse-Scale MachinesSynthesis Lectures on Computer Architecture, 2009
- Multilayer feedforward networks are universal approximatorsNeural Networks, 1989
- Logistic Regression, Survival Analysis, and the Kaplan-Meier CurveJournal of the American Statistical Association, 1988
- Induction of decision treesMachine Learning, 1986
- Maximum matching and a polyhedron with 0,1-verticesJournal of Research of the National Bureau of Standards Section B Mathematics and Mathematical Physics, 1965
- The Hungarian method for the assignment problemNaval Research Logistics Quarterly, 1955