TY - GEN
T1 - Network-aware program-counter-based disk energy management
AU - Crk, Igor
AU - Gniady, Chris
PY - 2009
Y1 - 2009
N2 - Reducing energy consumption is critical to prolonging the battery life in portable devices. With rising energy costs and increases in energy consumption by devices in stationary systems energy expenses of large corporations can annually reach into millions of dollars. Subsequently, energy management has also become important for desktop machines in large scale organizations. While energy management techniques for portable devices can be utilized in stationary systems, they do not consider network resources readily available to stationary workstations. We propose a network-aware energy management mechanism that provides a low-cost solution that can significantly reduce energy consumption in the entire system while maintaining responsiveness of local interactive workloads. The key component of the system is a novel program-context-based bandwidth predictor that accurately predicts application's bandwidth demand for file server/client interaction. Our dynamic mechanisms reduce the decision delay before the disk is spun-up, reduce the number of erroneous spin-ups in local workstations, decrease the network bandwidth, and reduce the energy consumption of individual drives.
AB - Reducing energy consumption is critical to prolonging the battery life in portable devices. With rising energy costs and increases in energy consumption by devices in stationary systems energy expenses of large corporations can annually reach into millions of dollars. Subsequently, energy management has also become important for desktop machines in large scale organizations. While energy management techniques for portable devices can be utilized in stationary systems, they do not consider network resources readily available to stationary workstations. We propose a network-aware energy management mechanism that provides a low-cost solution that can significantly reduce energy consumption in the entire system while maintaining responsiveness of local interactive workloads. The key component of the system is a novel program-context-based bandwidth predictor that accurately predicts application's bandwidth demand for file server/client interaction. Our dynamic mechanisms reduce the decision delay before the disk is spun-up, reduce the number of erroneous spin-ups in local workstations, decrease the network bandwidth, and reduce the energy consumption of individual drives.
UR - http://www.scopus.com/inward/record.url?scp=65749112532&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=65749112532&partnerID=8YFLogxK
U2 - 10.1007/978-3-642-01203-7_1
DO - 10.1007/978-3-642-01203-7_1
M3 - Conference contribution
AN - SCOPUS:65749112532
SN - 9783642012020
T3 - Studies in Computational Intelligence
SP - 1
EP - 13
BT - Software Engineering, Artificial Intelligence, Networking and Parallel/Distributed Computing
A2 - Lee, Roger
A2 - Ishii, Naohiro
ER -