Trends in Internet infrastructure drive a shift toward service-based computing. Internet-based service providers are adding new data center capacity for a variety of services including Web hosting, application services, outsourced storage, electronic markets and other network services. In these large scale and rapidly evolved data centers, one of the most important challenges is energy and thermal management. Since a data center usually consists of thousands of servers, network, disk and tape devices, the energy demand is so significant that sometimes it requires rewiring the whole building in order to host a data center. Moreover, it also results in high electricity bills. In the mean time, all these components constantly generate electric and/or mechanic heat and, if not cool down effectively, can significantly increase the room temperature, which can result in malfunctions of many devices and even shutdown of the entire data center.
This project addresses the above problems. More specifically, we study the energy and thermal management for data centers by making the following research thrusts:
- Energy and Thermal Modeling: evaluate the energy and thermal characteristics in a data center. More specifically, our project shall measure and model the energy consumption and heat dissipation by different components and identify key components for energy saving and heat reduction.
- Dynamic Energy and Thermal Management: investigate dynamic integrated solutions to conserve energy and minimize both electric and mechanic heat while still providing acceptable performance. In particular, the project focuses on dynamic energy and thermal management of memory-storage hierarchy.
- Managing Energy-Performance Tradeoffs for Multi-threaded Applications [SIGMETRICS'07]
- DMA-Aware memory energy management [HPCA'06]
- Hibernator: Helping Disk Array Sleep through the Winter [SOSP'05]
- Power Aware Storage Cache Management [IEEE-TC]
- Dynamically Tracking Miss-Ratio-Curve for Memory Management [ASPLOS'04]
- Performance-directed energy management for main memory and disks [ASPLOS'04] [IEEE Micro Top Pick] [ACM-TOS]
- Power-aware Storage Cache Replacement Algorithms [ICS'04]
- Reducing Energy Consumption of Disk Storage Using Power-Aware Cache Management [HPCA'04]
- Yuanyuan (YY) Zhou (Professor)
- Sarita Adve (Professor)
- Weihang Jiang
- Xiaodong(Jerry) Li
- Qingbo Zhu (graduated)
Collaborators:
- Sanjeev Kumar (Intel)
- Pei Cao (Cisco)
Funding:
- NSF Small ITR
- UIUC Startup Grant