## A Dynamic Disk Spin-Down Technique for Mobile Computing (1996)

Citations: | 155 - 7 self |

### BibTeX

@INPROCEEDINGS{Helmbold96adynamic,

author = {David P. Helmbold and Darrell E. Long and Bruce Sherrod},

title = {A Dynamic Disk Spin-Down Technique for Mobile Computing},

booktitle = {},

year = {1996},

pages = {130--142}

}

### Years of Citing Articles

### OpenURL

### Abstract

We address the problem of deciding when to spin down the disk of a mobile computer in order to extend battery life. Since one of the most critical resources in mobile computing environments is battery life, good energy conservation methods can dramatically increase the utility of mobile systems. We use a simple and efficient algorithm based on machine learning techniques that has excellent performance in practice. Our experimental results are based on traces collected from HP C2474s disks. Using this data, the algorithm outperforms several algorithms that are theoretically optimal in under various worst-case assumptions, as well as the best fixed time-out strategy. In particular, the algorithm reduces the power consumption of the disk to about half (depending on the disk's properties) of the energy consumed by a one minute fixed time-out. Since the algorithm adapts to usage patterns, it uses as little as 88% of the energy consumed by the best fixed time-out computed in retrospect. 1 In...

### Citations

669 | The weighted majority algorithm
- LITTLESTONE, WARMUTH
- 1994
(Show Context)
Citation Context ...cativeweight algorithmic family that has been developed by the computational learning theory community. This family has a long history and excellent performance for a wide variety of on-line problems =-=[14, 15, 16, 17, 18, 19, 20]. Algorith-=-ms in this family receive as input a set of "experts," other algorithms which make predictions. On each trial, each expert makes a prediction. The goal of the algorithm is to combine the pre... |

456 | Scheduling for reducing cpu energy
- Weiser, Welch, et al.
- 1994
(Show Context)
Citation Context ...f the mobile computer can be turned off or placed in low power modes to conserve energy. Several researchers have even considered dynamically changing the speed of the CPU in order to conserve energy =-=[1, 2]-=-. We show that a simple algorithm for deciding when to power down the disk drive is even more effective in reducing the energy consumed by the disk than the best fixed time-out value computed in retro... |

314 | How to use expert advice
- Cesa-Bianchi, Freund, et al.
- 1997
(Show Context)
Citation Context ...cativeweight algorithmic family that has been developed by the computational learning theory community. This family has a long history and excellent performance for a wide variety of on-line problems =-=[14, 15, 16, 17, 18, 19, 20]. Algorith-=-ms in this family receive as input a set of "experts," other algorithms which make predictions. On each trial, each expert makes a prediction. The goal of the algorithm is to combine the pre... |

259 | Comparing algorithms for dynamic speed-setting on a low-power CPU
- Govil, Chan, et al.
- 1995
(Show Context)
Citation Context ...f the mobile computer can be turned off or placed in low power modes to conserve energy. Several researchers have even considered dynamically changing the speed of the CPU in order to conserve energy =-=[1, 2]-=-. We show that a simple algorithm for deciding when to power down the disk drive is even more effective in reducing the energy consumed by the disk than the best fixed time-out value computed in retro... |

256 | UNIX disk access patterns
- Ruemmler, Wilkes
- 1993
(Show Context)
Citation Context ...ed algorithms, as well as the (impractical) best fixed time-out and optimal algorithms. 5.1 Methodology We used traces of HP C2474s disks collected from April 18, 1992 through June 19, 1992 (63 days) =-=[22]-=-. We compare the share algorithm with several algorithms, including the 2-competitive algorithm, the randomized ( e e\Gamma1 )-competitive algorithm, (an approximation to) the best fixed time-out, and... |

246 |
Aggregating strategies
- VOVK
- 1990
(Show Context)
Citation Context ...cativeweight algorithmic family that has been developed by the computational learning theory community. This family has a long history and excellent performance for a wide variety of on-line problems =-=[14, 15, 16, 17, 18, 19, 20]. Algorith-=-ms in this family receive as input a set of "experts," other algorithms which make predictions. On each trial, each expert makes a prediction. The goal of the algorithm is to combine the pre... |

209 | Adaptive disk spindown policies for mobile computers
- Douglis, Krishnan, et al.
- 1995
(Show Context)
Citation Context ...ce, and is simple to implement. In trace-driven simulations, it performs better than all of the algorithms described above, and even conserves more energy than the best fixed time-out. Douglis et al. =-=[13]-=- have recently studied some incrementally adaptive disk spin-down policies. The policies they consider maintains a changing time-out value. Whenever the disk access pattern indicates that the current ... |

197 |
Competitive snoopy caching
- Karlin, Manasse, et al.
- 1988
(Show Context)
Citation Context ...e than twice the energy used by the optimal algorithm. An algorithm is called c-competitive or has a competitive ratio of c if it never uses more then c times the energy used by the optimal algorithm =-=[10, 11]-=-. So this natural algorithm is 2-competitive, and we will refer to it as the 2-competitive algorithm. It is easy to see that the 2-competitive algorithm has the best competitive ratio of all constant ... |

194 | Tracking the best expert - Herbster, Warmuth - 1998 |

176 | Thwarting the powerhungry disk
- Douglis, Krishnan, et al.
- 1994
(Show Context)
Citation Context ...e algorithm for deciding when to power down the disk drive is even more effective in reducing the energy consumed by the disk than the best fixed time-out value computed in retrospect. Douglis et al. =-=[3]-=- show that the disk sub-system on portable computers consumes a major portion of the available energy (Greenawalt [4] states 30% or more). It is well-known that spinning the disk down when it is not i... |

158 | A quantitative analysis of disk drive power management in portable computers - Li, Kumpf, et al. - 1994 |

153 | Idleness is not sloth
- Golding, Bosch, et al.
(Show Context)
Citation Context ...b-system on portable computers consumes a major portion of the available energy (Greenawalt [4] states 30% or more). It is well-known that spinning the disk down when it is not in use can save energy =-=[3, 5, 6, 7]-=-. Since spinning the disk back up consumes a significant amount of energy, spinning the disk down immediately after each access is likely to use more energy than is saved. An intelligent strategy for ... |

126 | Empirical support for winnow and weighted-majority algorithms: results on a calendar scheduling domain
- Blum
- 1997
(Show Context)
Citation Context ...prove good worst-case bounds on the learning algorithm, the real-world performance of multiplicative weight algorithms appears less sensitive to the choice of parameters (for another example see Blum =-=[21]-=- on predicting calendar events). In our experiments, akin to a train and test regimen, we used a small portion of the data (the first day of one trace) to find a good setting for j and ff, and then us... |

115 |
Owickiâ€śCompetitive randomized algorithms for nonuniform problems
- Karlin, Manesse, et al.
- 1994
(Show Context)
Citation Context ...itive ratios. Although we still compute the competitive ratio based on a worstcase idle time between accesses, we average the energy used over the algorithm's random choice of time-out. Karlin et al. =-=[12]-=- give an (expected) ( e e\Gamma1 )-competitive randomized algorithm. If the spin-down cost is s, their algorithm chooses a time-out at random from [0; s] according to the density function p(time-out =... |

108 |
Mistake bounds and logarithmic linear-threshold learning algorithms
- Littlestone
- 1989
(Show Context)
Citation Context |

107 |
Learning when irrelevant attributes abound: a new linear-threshold algorithm
- Littlestone
- 1988
(Show Context)
Citation Context |

81 | Adaptive Disk Spindown via Optimal Rent-to-buy in Environments
- Krishnan, Long, et al.
- 1995
(Show Context)
Citation Context ...imer expires then the disk is spun down. This time-out can be set by the user, and typical values range from 30 seconds up to 15 minutes. Douglis et al. [3], Golding et al. [5], and other researchers =-=[4, 8, 6]-=- have proposed algorithms which spin the disk down more aggressively, conserving much more power than these relatively long time-outs. We use a simple algorithm called the share algorithm, a machine l... |

65 | Modeling power management for hard disks - Greenawalt - 1994 |

51 | Tight worst-case loss bounds for predicting with expert advice
- Haussler, Kivinen, et al.
- 1995
(Show Context)
Citation Context |

40 |
Predictive power conservation
- Wilkes
- 1992
(Show Context)
Citation Context ...b-system on portable computers consumes a major portion of the available energy (Greenawalt [4] states 30% or more). It is well-known that spinning the disk down when it is not in use can save energy =-=[3, 5, 6, 7]-=-. Since spinning the disk back up consumes a significant amount of energy, spinning the disk down immediately after each access is likely to use more energy than is saved. An intelligent strategy for ... |

19 |
Ousterhout, "Measurements of a Distributed File System
- Baker, Hartman, et al.
- 1991
(Show Context)
Citation Context ...ime-out, it is exploiting time dependencies in the input values. Of course, it is not surprising that there are time dependencies as it is well-known that user access patterns exhibit bursty behavior =-=[23]-=-. 250 300 350 400 450 500 550 600 650 700 5 10 15 20 25 30 35 40 Energy use during a 5 minute period Size of fixed time-out in seconds Fixed time-outs during a short interval Figure 6: Performance of ... |

15 |
Tarjan, "Amortized Efficiency of List Update and Paging Rules
- Sleator, E
- 1985
(Show Context)
Citation Context ...e than twice the energy used by the optimal algorithm. An algorithm is called c-competitive or has a competitive ratio of c if it never uses more then c times the energy used by the optimal algorithm =-=[10, 11]-=-. So this natural algorithm is 2-competitive, and we will refer to it as the 2-competitive algorithm. It is easy to see that the 2-competitive algorithm has the best competitive ratio of all constant ... |

1 |
Using experts for predicting continuous outcomes", in Computational Learning Theory: Eurocolt '93
- Kivinen, Warmuth
- 1993
(Show Context)
Citation Context |