Modeling and Analysis of the Thermal Properties Exhibited by Cyberphysical Data Centers

• Published in: IEEE systems journal Vol. 11; no. 1; pp. 163 - 172
• Main Authors: Malik, Saif U. R., Bilal, Kashif, Khan, Samee U., Veeravalli, Bharadwaj, Keqin Li, Zomaya, Albert Y.
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.03.2017; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Analytical models; Cloud computing; Computational modeling; cyberphysical systems (CPSs); data center (DC); Data centers; formal methods; Heating; modeling; Servers; Temperature; Thermal analysis; Thermal management; verification
• ISSN: 1932-8184; 1937-9234
• DOI: 10.1109/JSYST.2015.2493565

Data centers (DCs) contribute toward the prevalent application and adoption of the cloud by providing architectural and operational foundation. To perform sustainable computation and storage, a DC is equipped with tens of thousands of servers, if not more. It is worth noting that the operational cost of a DC is being dominated by the cost spent on energy consumption. In this paper, we model a DC as a cyberphysical system (CPS) to capture the thermal properties exhibited by the DC. All software aspects, such as scheduling, load balancing, and all the computations performed by the devices, are considered the "cyber" component. The supported infrastructure, such as servers and switches, are modeled as the "physical" component of the CPS. We perform detailed modeling of the thermal characteristics displayed by the major components of the CPS. Moreover, we propose a thermal-aware control strategy that uses a high-level centralized controller and a low-level centralized controller to manage and control the thermal status of the cyber components at different levels. Our proposed strategy is testified and demonstrated by executing on a real DC workload and comparing it with three existing strategies, i.e., one classical and two thermal-aware strategies. Furthermore, we also perform formal modeling, analysis, and verification of the strategies using high-level Petri nets, the Z language, the Satisfiability Modulo Theories Library (SMT-Lib), and the Z3 solver.