Energy consumption for data centers (a.k.a. hosting centers, server farms, clusters etc.) is rapidly increasing and is fast becoming a significant portion of the nation?s annual energy budget. Surprisingly, many of the contemporary data centers are designed and managed very inefficiently, mainly due to reliance on folklore techniques rather than those grounded in hard scientific evidence. In addition, contemporary research and experimentation in greening data centers is severely hindered by the unavailability of an experimentation test-bed, the cost of performing live tests of alternative configurations and the excessive long duration for performing simulations based on high-complexity offline models based on computational fluid dynamics. The BlueTool project aims to resolve these issues by: 1) increasing awareness of the latest scientific and engineering research on managing data centers, and 2) providing a research and evaluation infrastructure to test and develop new methodologies to address the inefficiencies of data centers. BlueTool will promote the use of holistic cyber-physical concepts to foster the development of energy-efficient and sustainable data centers. It will leverage recent research advances in cooling technologies and cyber-physical management for data centers at Arizona State University (ASU) and provide synergy for advancing the ongoing research in this field at ASU and elsewhere. BlueTool will consist of: (a) an online tool that can simulate various configurations of data centers, in terms of physical layout, hardware and software configuration; (b) a research hub and portal, maintained by the IMPACT Lab at ASU (http://impact.asu.edu), that offers data services on various updatable archives including power and thermal profiles, multi-scale low-complexity thermal and power models, and data center management methods and software. Researchers from both academia and industry will be able to use BlueTool's online consulting services to improve the computing performance and energy consumption of conventional or existing configurations with newly developed techniques.
Data centers (a.k.a. hosting centers, server farms) are rooms that can be as big as warehouses that contain hundreds or thousands of connected computers called servers. Many daily activities have to do with data centers, from shopping and making phone calls, to banking and web searching. Data centers are collectively one of the leading consumers of energy in the United States, accounting for about 3% of the US’ energy budget. There is a lot of research on how to make data centers energy-efficient. However, many of these solutions haven’t been validated because it takes an entire data center to test any of those ideas---and that is very expensive and not easy to do. The BlueTool project’s overarching goal was to make it easy for data center researchers to validate their energy-efficiency solutions in a manner in which both researchers and practitioners can benefit from the infrastructure. The BlueTool infrastructure offers the following facilities: i) BlueCenter: A small-scale experimental data center, geared with sensors, which is the heart of the BlueTool project. ii) BlueSense: A data center assessment toolkit, which consists of wireless portable sensor nodes for on-site thermal assessment of data centers. iii) BlueSim: A simulation toolkit that provides the means to verify models and management algorithms through detailed CFD simulation. iv) BlueWeb: A repository of information and the web presence of BlueTool. http://impact.asu.edu/BlueTool/. See Figures 1 and 2 for the structure of the project. The main teaching of BlueTool is that Data Centers should be viewed as cyber-physical systems (CPS). A CPS is any computer system which has dominant physical interactions with its immediate environment, in terms of movement, material exchange or energy exchange. Considering and modeling a data center as CPS, we can link quantitatively the amount of computational workload to how much energy is needed, what the heat conditions are in the data center room, and how much energy is wasted from the various components (see Figure 3). BlueSim has been used by simulation tool called GDCSim that can simulate various configurations of data centers, in terms of physical layout, hardware and software configurations. Major features of GDCSim include: 1) automated processing, operation without user intervention (especially the necessity of expert knowledge), 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, enabling design time testing and analysis of different data center configurations before deployment, 4) thermal analysis capability, which characterizes the thermal effects within the data center, 5) workload and power management, which enables workload scheduling and controlling the power modes of servers in the data center, and 6) consideration of cyber-physical interdependency, which enables feedback of information on temperature and air flow patterns in the data center to the management algorithms. BlueTool has helped: - In model validation: In validating the "heat recirculation matrix", a model that predicts temperatures given heat output from each server: In producing power curves, i.e. utilization-to-power models. BlueTool’s infrastructure was used to run experiments and yield what is called power curve. On developing a model for the computing-cooling trade-off effect, i.e. a condition where when consolidating the computational workload to fewer servers to save energy makes the cooling infrastructure work harder because of gathering all the heat to one side of the room. In research on models for operating data centers with energy storage. - Overall research in data centers: On how heat-activated cooling can be a feasible option forcooling data centers. This means that no significantly additional power is needed to run air conditioners. On the energy-wastage effects of phase imbalance of state-of the-art power distribution units. In reconsidering an assumption that power consumption in sleep mode can be assumed to be zero. On investigating the cooling delay effect, i.e., on the time it takes for extra cool air to reach the servers once they suddenly start heating up because of increased incoming workload. In students’ research and thesis work across the USA and abroad. - In education In teaching students about the cyber-physical nature of data centers. Several undergraduate and graduate students have used BlueTool in their research projects. The BlueTool team has made presentations to the research and business community about how cyber-physical models can help achieve considerable savings. The two main ways one can use BlueTool is to i) use the GDCSim and BlueSim simulators, ii) access Bluecenter and run an experiment, iii) consult the BlueTool team for any expert services. To access the services, visit the BlueTool home page and follow the links to the simulators (see Figure 4). Use BlueCenter (Figure 5): Contact the BlueTool team for obtaining a login account. You can then reserve a time to run your experiments. For more information and on how BlueTool can help you, email the BlueTool team at bluetoolteam@impact.asu.edu.
