Energy-Efficient Storage Systems

 

Students

Collaborators

We are at an interesting crossroads in computer architecture. Over the past two decades, there have been several advances in processor design, from microarchitectural techniques to boost instruction- and thread-level parallelism, to the shift from the use of single-core to multicore processors. The number of cores on the die is expected to continue growing in the future, thereby providing a tremendous amount of processing power within a single chip. At the same time, applications are also becoming more data intensive. There already exist several applications that handle massive amounts of data and are used by millions of people every day. These include traditional data intensive applications, such as transaction processing, email, and search engines as well as newer applications spanning areas such as social networking, photo and video sharing. The number of such applications is growing all the time and so is the amount of data they handle. This confluence of trends in architecture and applications requires computer architecture support to facilitate the storage of massive amounts of data and providing efficient access to them and also to efficiently transport the data between storage and the processors. Designing such computer architectures in an energy-efficient manner is a major challenge.

The goal of this project is to design energy-efficient storage systems and storage-centric architectures for data intensive applications, as well as develop techniques for designing and configuring storage systems to boost energy efficiency. A few contributions of this project to date include Dynamic RPM modulation to provide multiple performance/energy operating points for disk drives, Intra-Disk Parallelism to exploit I/O parallelism inside a disk drive, sensitivity-based optimization to craft storage energy management policies, and active storage architectures that facilitate storage-centric computation to be offloaded to processors closer to the storage devices.

Representative Publications