Quantitative Evaluation of Systems, International Conference on (2006)
Sept. 11, 2006 to Sept. 14, 2006
DOI Bookmark: http://doi.ieeecomputersociety.org/10.1109/QEST.2006.15
Arif Merchant , Hewlett-Packard Laboratories, Palo Alto, CA
Modern data centers can contain thousands of hosts and hundreds of disk arrays, with total storage capacities approaching petabytes. The drive for reducing the cost of the IT infrastructure leads to consolidation of the enterprise's applications, and a single data center may support thousands of applications, with users spread over continents and administrated by IT personnel with no direct connection with the users of the application. In a conventional IT environment, applications are usually managed by a small group of administrators who understand the application's requirements and how it interacts with the hardware and storage. This approach breaks down in the large data center, where the administrators may manage many applications and have only a limited understanding of their behavior. Moreover, because of the degree of resource sharing in a data center, the interactions between applications are many and complex, and it is difficult for the administrators to predict the interactions well enough to generate good designs; as a result, the storage solutions are often over-engineered.
A. Merchant, "Designing and managing storage systems: issues, techniques, and challenges," Quantitative Evaluation of Systems, International Conference on(QEST), Riverside, California, 2006, pp. 265-268.