Energy efficiency has become a big issue in both data centers, mobile devices, and embdded systems. Over the past five years, there has been a significant increase in the number of data centers, in addition to an estimated doubling of energy used by servers and cooling systems. According to the EPA, the U.S. data centers will consume 100 billion kilowatt hours annually by 2011. The total cost for the power and cooling for servers in the US is expected to reach $45B in 2010. Moreover, the emergence of new technologies and a wide range of new mobile revolutionary products and applications have introduced a need for more power resources in mobile devices. Many improvements have been done, however, it can be noticed that the improvements of the existing power resources are limited and still do not meet the requirement of longer battery life time. Therefore, improving energy-efficiency in both data centers and mobile devices has become more and more demanding. The key challenges are reducing carbon emissions and saving cost. Here is a brief introduction of sustainable computing.
The mission of LaST (Laboratory for Sustainable Computing) is investigating the enabling technologies for sustainable computing, including energy-aware operaring systems, intelligent resource scheduling and management, community shared computing, thermal management for multicore. The long term goal of this lab is to reduce the energy consumption of computing devices significantly and/or enforce the thermal management in real-time fashion, while utilizing energy more efficiently. Specifically, the three on-going projects are:
Understand the energy consumption of applications in a computer system in a real time fashion. We intend to provide a tool that will answer the question of how much energy is consumed for each resource, such as CPU, memory, disk and network, by each application in real time at a fine-grained level. This tool will enable other optimizations, from the perspective of both the system and user(s). The source code of pTop is free downloadable at sourceforge.net. We recently released the windows of pTop, pTopW.
Design and implement efficient power management schemes for mobile devices and large data centers. As the Internet has increasingly headed into the new cloud computing era, we envision that power management schemes from both the users’ side (mobile devices) and server side (data centers) need to be investigated. We plan to attack these two scenarios from different angles.
Thermal Stress Analysis for Real-Time Systems .We intend to formally incorporate a system’s real-time requirements into dynamic thermal management (DTM) mechanisms: we propose to address this objective by developing a framework for thermal-stress analysis of real-time systems which includes methodology, analysis, and tools required to design and implement real-time systems that behave predictable under unexpected thermal changes in the environment.