Designing for disasters

Kimberly Keeton, Cipriano Santos, Dirk Beyer, Jeffrey Chase, John Wilkes
Hewlett-Packard Laboratories, Palo Alto, CA and Duke University, Durham, NC

{kimberly.keeton, cipriano.santos, dirk.beyer, john.wilkes}@hp.com, chase@cs.duke.edu

Losing information when a storage device or data center fails can bring a company to its knees—or put it out of business altogether. Such catastrophic outcomes can readily be prevented with today’s storage technology, albeit with some difficulty: the design space of solutions is surprisingly large, the configuration choices are myriad, and the alternatives interact in complicated ways. Thus, solutions are often over- or under-engineered, and administrators may not understand the degree of dependability