Online Updating and Scheduling of Computer Models with Application to Data Center Thermal Management

• Activity Number: 533
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 10:30 AM to 12:20 AM
• Sponsor: Section on Physical and Engineering Sciences
• Author(s): Huijing Jiang*+ and Xinwei Deng and Vanessa Lopez and Hendrik F. Hamann
• Companies: IBM T.J. Watson Research Center and Virginia Tech and IBM T.J. Watson Research Center and IBM T. J. Watson Research Center
• Keywords: computer model ; dynamic bias correction ; Kalman filter ; stochastic projected gradient methods ; data center enery management

Abstract

Energy and thermal management in data centers has become very important not only because of raising cost of energy but also in order to address growing sustainability concerns about information technology. Towards this end, complex physics-based computer models (e.g., computational fluid dynamics models) are extensively used to simulate the temperature distribution map in data centers. However, due to management policies and time constraints, it may not always be practical to use such models in real-time. In this talk, we propose a novel statistical modeling approach to perform real-time simulation by dynamically correcting the discrepancy between a base, steady-state solution of a physical model and real-time temperature sensor data. The proposed method utilizes Kalman filter and stochastic gradient descent approach as computational tools to achieve real-time updating of the base temperature map as new sensor measurements are acquired. We evaluate the performance of our proposed method through a simulation study and demonstrate its merits in a case study conducted for a real-life data center.