SOLVING THE ENERGY CRISIS FOR THE DATA CENTER
Computer Science Department
Electrical and Computer Engineering Department
University of Pittsburgh
210 S. Bouquet St, Pittsburgh, PA 15260 USA
Point of contact: Bruce Childers (childers "at" cs.pitt.edu)
NEWS
Recent Talks by the PIs
11/07/11: Pitt Chronicle article about PCM@Pitt
Joint Computer Science and Electrical and Computer Engineering faculty awarded an NSF grant to study Energy Efficient Data Centers -- CSR: Large: Storage Class Memory Architecture for Energy Efficient Data Centers (National Science Foundation, award CSR/CNS 1012070). A group of Computer Science and Electrical and Computer Engineering faculty, including Bruce Childers (PI), Sangyeun Cho (Co-PI), Rami Melhem (Co-PI), Daniel Mosse' (Co-PI), Jun Yang (Co-PI), and Youtao Zhang (Co-PI) are working together to define a new computer memory system for data centers that has exceptionally low power consumption . This project uses multiple technologies to construct a high-capacity, energy efficient memory system for virtualized computer servers with a new Storage Class Memory Architecture that incorporates multiple memory technologies such as DRAM, Phase-change memory (PRAM) and Flash.
This project is part of the PittCS Sustainable Computing effort.
OVERVIEW
Publications
By 2011, data centers in the U.S. are poised to consume around 100 billion KW of energy, half of which in servers. The majority of power consumption will shift from the processor to main memory. This project, rather than relying solely on DRAM, uses multiple technologies to construct a high-capacity, energy-efficient memory system for virtualized computer servers with a new Storage Class Memory Architecture that incorporates multiple memory technologies such as DRAM, Phase-change memory (PRAM) and Flash.
This project is ambitious with contributions in the design and development of energy-efficient computer servers. The technical contributions include: a new integrated memory architecture that seamlessly manages hybrid memory resources; novel algorithms and policies for energy, performance and endurance management; better understanding of the hypervisor's role in resource management and its interaction with upper and lower system layers; and a new hybrid DRAM and PRAM main memory controller. The infrastructure contributions are in the form of tools that will be developed and disseminate, including simulations, analytic models and actual software/emulated hardware system.
This research represents a new "green computing technology" that satisfies the demands of future computing systems (e.g., cloud computing, server farms, commodity computers) in an exceptionally low energy manner. The project impacts undergraduate and graduate students through research opportunities and industry internships and aims to recruit people from under-represented groups in computer science and computer engineering.