#### DMCA

## Algorithms for power savings (2003)

Venue: | In SODA ’03: Proceedings of the fourteenth annual ACM-SIAM symposium on Discrete algorithms |

Citations: | 125 - 6 self |

### Citations

542 | A scheduling model for reduced CPU energy
- Yao, Demers, et al.
- 1995
(Show Context)
Citation Context ...hat enables the sensors. Data can be transmitted at di erent speeds and each speed has a di erent power usage rate. The system also has a sleep state in which thepower usage level is greatly reduced. =-=[16, 14]-=- 2 Previous Work The problem of when to transition a device to a sleep state when it is idle is a continuous version of the Ski Rental problem [5]. It is well known that the optimal competitive ratio ... |

390 | Voltage scheduling problem for dynamically variable voltage processors
- Ishihara, Yasuura
- 1998
(Show Context)
Citation Context ... system can run and the algorithm must select one of these values. Some work in the systems literature address models in which the system can not change instantaneously or continuously between speeds =-=[3, 4]-=-. Naturally, this makes the problem much harder to solve. As a result, much ofthe work on Dynamic Speed Scaling makes all of the assumptions we make here. It remains to determine experimentally whethe... |

163 | Speed scaling to manage energy and temperature
- Bansal, Kimbrel, et al.
- 2007
(Show Context)
Citation Context ...a degree-d polynomial. Recently, Bansal, Kimbrel and Pruhs have shown that another natural online algorithm called Opitmal Available obtains a competitive ratioofd d which is tight for that algorithm =-=[1]-=-. They also introduce a new algorithm which obtains a competitive ratio of 2(d=(d ; 1)) d e d . Dynamic Speed Scaling is also a well studied problem in the embedded systems literature. (See [11] and r... |

49 | Online strategies for dynamic power management in systems with multiple powersaving states - Irani, Shukla, et al. - 2003 |

47 | Competitive Analysis of Dynamic Power Management Strategies for Systems with Multiple Power Saving States
- Irani, Shukla, et al.
- 1996
(Show Context)
Citation Context ...le period is generated by aknown probability distribution [8]. Irani et al. examine the generalization in which there are multiple sleep states, each with a di erent power usage rate and startup cost =-=[6]-=-. There has also been experimental work that investigates how to use trace data to estimate a probability distribution that can be used to guide probabilistic algorithms [9, 6]. The embedded systems l... |

38 | An empirical evaluation of virtual circuit holding time policies in IP-over-ATM networks
- Keshav, Lund, et al.
- 1995
(Show Context)
Citation Context ...sage rate and startup cost [6]. There has also been experimental work that investigates how to use trace data to estimate a probability distribution that can be used to guide probabilistic algorithms =-=[9, 6]-=-. The embedded systems literature refers to the problem of deciding when to transition to a low-power sleep state as Dynamic Power Management. Benini, Bogliolo and De Micheli give an excellent review ... |

32 | Low-power task scheduling for multiple devices
- LU, BENINI, et al.
- 2000
(Show Context)
Citation Context ... for an embedded system called SmartBadge [15]. Another related paper examines task scheduling (although not with multiple speeds) so as to create idle periods for putting a device into a sleep state =-=[10]-=-. This problem captures some of the features of the problem we address here. There are a number of issues in the real-world problem of power management that are not incorporated into the model we use ... |

14 |
Randomized competitive algorithms for non-uniform problems
- Karlin, Manasse, et al.
- 1990
(Show Context)
Citation Context ... that can be achieved by any deterministic online algorithm for this problem is 2. Karlin et al. examine the problem when the length of the idle period is generated by aknown probability distribution =-=[8]-=-. Irani et al. examine the generalization in which there are multiple sleep states, each with a di erent power usage rate and startup cost [6]. There has also been experimental work that investigates ... |

12 |
De Micheli “A Survey of Design Techniques for System-Level Dynamic Power Management
- Benini, Bogliolo, et al.
- 2000
(Show Context)
Citation Context ...d systems literature refers to the problem of deciding when to transition to a low-power sleep state as Dynamic Power Management. Benini, Bogliolo and De Micheli give an excellent review of this work =-=[2]-=-. The Dynamic Speed Scaling problem without the sleep state has been examined by Yao, Demers and Shenker (although not under that name) [17]. They give an optimal o ine algorithm for the problem. Thei... |

4 |
Online Computation.” From Approximations for NP-Hard Problems
- Irani, Karlin
(Show Context)
Citation Context ...n which thepower usage level is greatly reduced. [16, 14] 2 Previous Work The problem of when to transition a device to a sleep state when it is idle is a continuous version of the Ski Rental problem =-=[5]-=-. It is well known that the optimal competitive ratio that can be achieved by any deterministic online algorithm for this problem is 2. Karlin et al. examine the problem when the length of the idle pe... |

1 |
Energy E cient Fixed-Priority Scheduling for Real-Time Systems on Variable Voltage Processors
- Quan, Hu
- 2001
(Show Context)
Citation Context ...orithm [1]. They also introduce a new algorithm which obtains a competitive ratio of 2(d=(d ; 1)) d e d . Dynamic Speed Scaling is also a well studied problem in the embedded systems literature. (See =-=[11]-=- and references therein). The problem often goes by the name Dynamic Voltage Scaling or Dynamic Frequency Scaling. We adopt the more generic term Dynamic Speed Scaling to emphasize the fact that the a... |

1 |
Adaptive power- delity in energy aware wireless embedded systems
- Raghunathan, Spanos, et al.
- 2001
(Show Context)
Citation Context ... prove two results in this paper. These results hold for convex power functions P (s). The convexity of P (s) is a standard assumption in this area and corresponds well to analytical models for P (s) =-=[12]-=-. We give an o ine algorithm for the DSS-S problem that produces a schedule whose total energy consumption for any set of jobs is within a factor of two of optimal. We still do not know whether the o ... |

1 |
Latency E ects of System Level Power Management Algorithms," 2000
- Ramanathan, Irani
(Show Context)
Citation Context ...porate the latency involved in transitioning from the on to the sleep state and vice versa [2]. Ramanathan et al. perform an experimental study of the latency/power tradeo in Dynamic Power Management =-=[13]-=-. In [6], algorithms which are designed using a model which does not incorporate this latency perform very well empirically even when this additional latency is taken into account. The model we use he... |

1 |
Energy E cient System Design and Utilization
- Simunic
- 2001
(Show Context)
Citation Context ...lgorithm selects the speed of the system to minimize power usage. Simunic examines the problem of combining Dynamic Speed Scaling and Dynamic Power Management for an embedded system called SmartBadge =-=[15]-=-. Another related paper examines task scheduling (although not with multiple speeds) so as to create idle periods for putting a device into a sleep state [10]. This problem captures some of the featur... |