The Case for Storage-Centric Architecture DesignReport
The proliferation of digital data is likely to cause storage and information-centric applications to be a key part of future enterprise workloads. This likely change in the workload mix will need a corresponding rethinking of system architecture designs to be storage-centric. In this paper, we evaluate one possible dimension in this space: the offload of computation to processing elements closer to the storage. We make two contributions in this regard. First, we examine a continuum of choices where the computation is offloaded to existing system components–a disk drive processor, a disk array controller, or a management processor. Our evaluation demonstrates that there are interesting tradeoffs in the choice of each location. Second, we evaluate the impact of microarchitectural changes to enhance the offload processor. We show that our new architecture can achieve orders of magnitude higher performance, at significantly lower power.
Note: Abstract extracted from PDF text
All rights reserved (no additional license for public reuse)
Tarapore, Shahrukh, Clinton Smullen, Sudhanva Gurumurthi, Parthasarathy Ranganathan, and Mustafa Uysal. "The Case for Storage-Centric Architecture Design." University of Virginia Dept. of Computer Science Tech Report (2007).
University of Virginia, Department of Computer Science