|
|
|
||
Applications such as Facebook, Gmail, and Flickr must store terabytes of data, process hundreds of thousands of user requests every second, achieve acceptable service quality, and protect users' data from any corruption or loss. In addition, storage is a major driver of datacenter costs and can consume significant amounts of energy, even during periods of low demand. Thus, there is a need for storage systems that deliver high capacity, predictable (but not necessarily minimal) latencies, and low power demand, all for reasonable installation costs. I am using analytic performance modeling techniques to investigate designs for new storage sytems in the cloud. Some of the problems I am studying include tradeoffs between conventional hard disks and solid-state drives, the design of optimal storage architectures, and design to meet service-level objectives. |
