Modified on 2010/11/25 19:21 by Administrator — Categorized as: Uncategorized

Cyber-physical systems (CPS) are physical and engineered systems whose operations are monitored, coordinated, controlled and integrated by a computing and communication core. Data centers with their voracious appetite for power and cooling are an embodiment of cyber-physical systems that require ambient intelligence. Energy-efficient operations and the integrated management of cyber- and physical aspects of data centers are the central focus of this proposed effort.

Challenges faced by Data Centers

Data centers increasingly constitute a critical backbone of the worldwide IT infrastructure, serving as the server infrastructure for search engines, mail servers, e-commerce, data warehousing and other cloud computing functions. IBM, for example, operates more than 400 data centers across the world occupying more than 7 million square feet. While such data centers generally target large-scale virtual IT services, the design, construction and operation of data centers (a) depend on cyber-physical infrastructure with major power and cooling requirements, (b) incur significant energy costs, and (c) can lead to significant economic and societal impact from the failures of physical subsystems. In fact, power and cooling in a data center cost more than the information technology (IT) equipment supported. As a result, data centers face an emerging crisis. The challenges faced by these data centers include:
This project aims at tackling some of these challenges to to enable Sustainable ENergy-Optimized Datacenters (SENODs).

Expected Benefits

Significant reduction in electricity and cooling costs are targeted at each data center. If even 10% or more improvements in equipment density can be made safely, the construction of new data centers can be deferred. The lack of good energy models can lead to lower operating temperatures than specified. An integrated toolset that couples physical and cyber information will lead to significantly better thermal models. Precise knowledge of the difference between the actual power used and the power allocated can offer opportunities for increasing the density of equipment. This will yield additional energy- and cost-efficiencies. Dynamic layout optimization as the data center evolves will also make it more energy-efficient. Failures of physical subsystems that can endanger the health of the data center will be detected and notified. Our fine-grained sensing may also help to detect any minor problems before they become too large or very time-critical. Intrusions and associated ill-effects will be minimized. Accurate energy consumption reports can be provided to enterprise customers using the data centers. Simplified operations will also reduce operator errors.