Overview
The Energy Management & Sensor Networking group at IBM Research - Zurich is engaged in R&D projects related to the energy management of data centers (DCs) and buildings, with a focus on real-time sensing technologies. We evaluate wired and wireless sensor networks at the physical, medium access and networking level, including publish–subscribe messaging and its integration into IBM's middleware application framework, such as WebSphere. We pursue prototyping and testing with sensing applications in realistic DC environments using a variety of both commercial and experimental variants of standardized as well as proprietary wireless sensor networks.
Challenge of data centers
The rising energy consumption of DCs has emerged as a key challenge for the IT industry. For example, if a DC cannot accommodate additional servers or storage racks because of insufficient power supply, inefficient cooling or other infrastructure constraints, the underlying inefficiencies can impede the ability to grow IT services. According to an EPA research report, the energy consumption of IT equipment and DCs in the USA will nearly double in the next five years [1]. Similar developments are expected in other geographies.
Generally, about half of the energy supplied to DCs is consumed by the associated cooling infrastructures. Among its Big Green initiatives, IBM is developing technologies dedicated to improve the efficiency of DCs and thus to reduce the costs associated with energy use and cooling needs. The combination of wired and/or wireless sensor technologies with energy management software provides real-time monitoring and energy management capabilities for DCs as well as building facilities. Customers can identify inefficiencies in DC operations and initiate corrective actions in a timely fashion to reduce energy consumption and to improve both the efficiency and the reliability of their DC.
IBM's R&D teams are pursuing a systematic approach to determine the operating conditions in a DC with a dedicated measurement and management technology (MMT) [2, 3] that provides precise information on DC energy consumption and temperature hot-spots. High-performance servers can create hot-spots that excessively burden DC cooling systems, making it difficult to keep energy consumption within desirable bounds. Thus, it is critical to obtain accurate and detailed information on the environmental and operating conditions in a DC to be able to identify opportunities to improve energy efficiency.
