GDCSim - An Integrated Tool Chain for Analyzing Green Data Center Physical Design and Resource Management Techniques Sandeep K.S. Gupta, Rose Robin Gilbert, Ayan Banerjee, Zahra Abbasi, Tridib Mukherjee, Georgios Varsamopoulos Impact Lab, School of Computing, Informatics and Decision Systems Engineering, ASU, Tempe, Arizona {sandeep.gupta,roserobin.gilbert,abanerj3,zabbasi1,tmukherj,georgios.varsamopoulos}@asu.edu Abstract—Energy consumption in data centers can be reduced by efficient design of the data centers and efficient management of computing resources and cooling units. A major obstacle in the analysis of data centers is the lack of a holistic simulator, where the impact of a new computing resource (or cooling) management approach can be tested with different design (i.e., layout and configuration) of data centers (or vice versa). To fill this gap, this paper proposes Green Data Center Simulator (GDCSim) for analyzing data center energy efficiency by studying and testing: (i) different data center geometries, (ii) workload characteris- tics, (iii) platform power management schemes; (iv) scheduling algorithms; and (v) data center configurations. GDCSim is used to iteratively design green data centers. Further, it is validated against established CFD simulators. GDCSim is developed as a part of the BlueTool infrastructure project at Impact Lab. Keywords-Simulator; energy efficient; green, data center; I. I NTRODUCTION Green initiatives in data center design encompasses energy efficiency, thermal awareness and usage of renewable sources of energy. To achieve energy efficiency and thermal awareness, data center designers have mainly focused on two primary objectives: (a) improving physical design [30], which includes reducing heat recirculation and hence hot-spots (to reduce the cooling demand), providing efficient cooling mechanisms [3] (to reduce cooling energy consumption), and developing low power computing equipment [15], [18]; and (b) developing energy-efficient computing resource management strategies, such as workload scheduling algorithms [24], [28], [32], cool- ing management [5], and power management schemes [21]. Conventional methodologies for analyzing green strategies in data centers have considered: 1) energy efficiency analysis of resource management algorithms, which include workload scheduling [12], [17], [20], active server set selection [12], [31], and power management of servers with dynamic voltage and frequency scaling [4], 2) thermal analysis of the data center under different physical configurations and resource management strategies [6], [7], [19], [30], and 3) closed loop operation of the resource management strategies with the physical behavior such as thermal behavior and air flow within the data center [9], [31], [32]. Of much interest is the close operation of the resource management algorithms and physical behavior in a data center through thermal interactions. For example, the extent of heat recirculated in a data center (i.e., a physical design attribute) may affect the energy efficiency of a workload scheduling algorithm [28] (i.e., the computing operation). Similarly, the workload placement may cause hot spots and thus demand a redesign of the data center lay- out [19]. It is thus imperative to consider such cyber-physical inter-dependencies for devising techniques for the two primary objectives for green data center design. A holistic design of green data center, will consequently require application of all such techniques in a unified analysis framework. This paper proposes Green Data Center Simulator (GDCSim),a simulation tool that unifies existing techniques to green data center management and allows holistic design and analysis before deployment. Such a holistic simulation tool should have the following key features as shown in Table I. The key features include: 1) automated processing, which allows interfacing different modules together and does not require user intervention once the simulation has started, 2) online analysis capability, which allows real time simulation of management decisions based on changes in the physical environment in the data center, 3) iterative design analysis, which enables design time testing and analysis of different configurations before deployment, 4) thermal analysis capability, which characterizes the thermal effects within the data center room at a given time, however, does not include thermal feedbacks to the management of the data center, 5) workload and power management, which enables workload scheduling and controlling the power modes of servers for higher data center efficiencies, and 6) con- sideration of cyber-physical interdependency, which enables feedback of information on temperature and air flow patterns in the data center to the management algorithms and the closed loop operation of the servers and cooling units (CRAC) to achieve energy efficient operation. From Table I, it can be seen that considerable amount of research is devoted to online analysis of workload management algorithms which also consider the cyber-physical inter-dependencies within the data center. Further, Computational Fluid Dynamics (CFD) simulator tools to characterize the thermal effects and airflow patterns also exists in abundance. However a simulator, which enables holistic design of a data center with the above features is missing. GDCSim, proposed in this paper, fills this gap. The principal obstacle, in this regard, is an accurate charac- terization of the cyber-physical inter-dependencies. Accurate characterization of the physical behavior of the data center involves CFD simulations, which are extremely slow and time consuming. On the other hand, resource management strate- gies, often use feedback from physical world to take online decisions on placement of workloads, cooling adjustment and power management of servers. Analyzing such algorithms require, online testing of a resource management strategy on 978-1-4577-1221-0/11/$ 26.00 c 2011 IEEE 1