John Wilkes. Predictive Power Conservation. Technical Report HPL-CSP-92-5, HewlettPackard, May 1992.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....to having perfect knowledge of the future and applying both direct deactivation and pre activation. Rather than simulate, we implement and experimentally evaluate direct deactivation and pre activation. Conserving disk energy. Disks have been a frequent focus of energy conservation research, e.g. [21, 12, 3, 2, 8, 15, 10, 6, 14]. The vast majority of the previous work has been on history based, adaptive threshold policies, such as the one used in IBM disks. Because our application supported policies can use information about the future, they can conserve more energy and avoid performance degradation more effectively than ....

J. Wilkes. Predictive Power Conservation. Technical Report HPL-CSP-92-5, Hewlett-Packard, May 1992.

....perfect knowledge of the future and applying both direct deactivation and pre activation. Rather than simulate, we model, implement, and experimentally evaluate direct deactivation and pre activation. Conserving Disk Energy. Disks have been a frequent focus of energy conservationresearch, e.g. [16, 11, 3, 2, 7, 14, 9, 6, 13]. The vast majority of the previous work has been on history based, adaptivethreshold policies, such as the one used in IBM disks. Because our application supported policies can use information about the future, they can conserve more energy and avoid performance degradation more effectively than ....

J. Wilkes. Predictive Power Conservation. Technical Report HPL-CSP-92-5, Hewlett-Packard, May 1992.

....consumed by not spinning down the disk, and about 40 compared to the common practice of spinning down the disk after 3 5 minutes of idle time. However, the fine grained approach causes several spin up delays per hour [6] Wilkes proposed the use of a predictive algorithm for disk management [21] by predicting and adjusting the spin down delay based on a weighted average of recent activity durations. Golding, et al. placed this proposal in a more general framework and studied a number of spindown delay prediction methods, including arithmetic and geometric adjustments of predicted ....

J. Wilkes. Predictive power conservation. Technical Report HPL-CSP-92-5, Hewlett-Packard Laboratories, Feb. 1992.

....5 explores the effectiveness of server side assistance in conserving the WNIC energy on the client. We conclude in Section 6. 2 Related work There has been considerable work on power management for components of a mobile device. This work includes spindown policies for disks and alternatives [35, 3, 26, 12, 19], scheduling policies for reducing CPU energy consumption [34, 17] and managing wireless communications [20, 11, 21, 30] Our work is similar in spirit to the work of Feeney et al. 15] They obtain detailed measurements of the energy consumption of an IEEE 802.11 wireless network interface ....

John Wilkes. Predictive Power Conservation. Technical Report HPL-CSP-92-5, HewlettPackard Labs, February 1992.

....facilitate power management. Fig. 15. Steps of a Linux scheduler. C. Predictive Wakeup If a requests requires a sleeping device, this request has to wait for the wakeup delay. Ideally, the device should wake up before the request arrives to eliminate waiting; this is called predictive wakeup [38]. A device should wake up just before requests arrive. Waking up too early wastes energy; waking up too late does not eliminate waiting. Our experiments show that the method in [16] actually increases energy because it mixes requests from all processes and has low prediction accuracy. To perform ....

J. Wilkes, "Predictive power conservation," Hewlett-Packard, Tech. Rep., HPL-CSP-92-5, 1992.

....took a broad approach to the effects of OS controlled power conservation, and found that revisiting techniques for application OS interaction, memory allocation, resource protection and allocation, and communication may have significant effects on energy consumption of future computers. Wilkes [Wil92] focuses on determining the proper time to spin down a hard disk, while [LT95] focuses specifically on memory allocation techniques to maximize simultaneous data transfers. 6 Conclusions and Future Work Conclusions This paper presents a model for determining the energy savings. This simplistic ....

John Wilkes. Predictive power conservation, February 1992.

....that could be used to balance these concerns, they provide no guidance as to how these parameters should be set. 8.2.3 Storage power management Most efforts to reduce storage energy consumption have concentrated on saving power by spinning down disk drives when they are not in use. Wilkes [98] first presented the idea of adaptive disk power management, which uses a history of past disk activity to predict when the disk can be productively put in a low power mode. He also hypothesized that a similar algorithm might be useful for determining when disks should be restored to full power ....

Wilkes, J. Predictive power conservation. Technical Report HPL-CSP-92-5, Hewlett-Packard Laboratories, February 1992.

....the modi cation was performed. Fig. 12. PowerPoint power usage with 1 minute autosave The OS saves energy by spinning down the hard drive if the disk remains in a low power state long enough to counterbalance the energy cost of power state transitions. Timeout based power management policies [3, 4, 9, 15] are based on the observation that disk accesses are often closely correlated together in time. Immediately after an access is observed, it is likely that another access will be seen soon. Spinning down the disk is undesirable, since the next access will likely occur before the break even point ....

Wilkes, J. Predictive power conservation. Technical Report HPL-CSP-92-5, Hewlett-Packard Laboratories, February 1992.

....highly unpredictable and hence offer little possibility for history based prediction mechanisms (attenuated by high threshold) 5 Related Work There has been considerable work on power management for components of a mobile device. This work includes spindown policies for disks and alternatives [4, 12, 11, 10, 16, 19, 21, 35], and managing wireless communication [17, 18, 33, 30, 9] Agrawal et al. 2] describe techniques for processing video data for transmission under low battery power conditions. Corner et al. 8] describe the time scales of adaptation for mobile wireless video conferencing systems. In this work, we ....

John Wilkes. Predictive Power Conservation. Technical Report HPL-CSP-92-5, Hewlett-Packard Labs, February 1992. 26

....available power [4] One common approach has been to exploit idle periods, either by shutting down the subsystem or by putting the idle time to productive use. Spinning the disk down or putting the display screen into a screen saving mode when it is not in use has been shown save significant power [4, 21]. Golding et al. 6] have also generalized the idle time processing setting, and provided a comprehensive taxonomy of idle time detection and prediction algorithms. This thesis first explores the wireless modem setting in detail to determine if idle prediction methods could be utilized. Since the ....

John Wilkes. Predictive power conservation. Technical Report HPL-CSP-92-5, Hewlett-Packard Laboratories, February 1992.

....in the literature. Douglis et al. DKM94] show that the disk sub system on portable computers consumes a major portion of the available energy (Greenawalt [Gre94] states 30 or more) It is well known that spinning the disk down when it is not in use can save energy [DKM94, GBS 95, LKHA94, Wil92] 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 deciding when to spin down the disk is needed to maximize the energy savings. Current mobile ....

....case assumptions. Researchers introduce new algorithms with provably good cumulative loss over a sequence of trials. 3.1 Research in disk spin down Several researchers have studied disk spin down for power conservation. Wilkes proposes that intelligent disk spin down algorithms can save power [Wil92] Douglis et al. instrument several computers, collect disk traces, and demonstrate on these traces that short fixed timeouts can dramatically improve power savings [DKM94] These early works point out the potential benefits of disk spin down algorithms. One of the simplest algorithms for disk ....

John Wilkes. Predictive power conservation. Technical Report HPL-CSP-92-5, Hewlett-Packard Laboratories, February 1992.

....or HTTP responses. All ICAP transactions are based on standard HTTP. 2.2.3 Battery power and storage constraints 2.2.3.1 Power Aware Systems There has been considerable work on power management for various system components. This work includes spindown policies for disks and alternatives [12, 36, 35, 34, 56, 75, 80, 121], and managing wireless communication [59, 74, 111, 107] Lorch et al. 83] present a survey of the various software techniques for energy management. We should be able to exploit power management techniques for the components upon which our system depends on. Unfortunately, our usage model ....

John Wilkes. Predictive Power Conservation. Technical Report HPLCSP -92-5, Hewlett-Packard Labs, February 1992.

....PADRAM class memory devices. A novel aspect of our work is the cooperative hardware OS approach to exploit PADRAM features. Previous OS level studies focusing on power management include work on scheduling for low power processor modes [31, 32, 46] spindown policies for disks and alternatives [1, 7, 8, 9, 16, 25, 30, 47], and managing wireless communication [18, 24, 42] A consortium of companies has developed a specification [20] that addresses the lower level OS device interface, providing one model for gross system wide power states and per component power states as a basis for the development of OS directed ....

J. Wilkes. Predictive Power Conservation. Technical Report HPL-CSP-92-5, Hewlett-Packard Labs, February

....trace driven simulation to look at a fixed threshold policy for disk spin control, and studied buffer cache parameters. Wilkes hypothesized that it would be effective to use a weighted average of a few previous interarrival times at disk to decide when to spindown the disk on a mobile computer [Wil]. Adaptive spindown policies that continually change the spindown threshold based on the perceived inconvenience to the user are studied in [DKB] Golding et al. GBS] have studied idle time detection and prediction, and proposed a taxonomy of idle time detection 6.3. OUR APPROACH 59 algorithms; ....

....it would be effective to use a weighted average of a few previous interarrival times to decide when to spindown the disk on a mobile computer. He noted as well that if inactive intervals were of roughly fixed duration, the disk could be spun up in advance of the expected time of the next operation [Wil]. If access patterns are not so consistent, however, these techniques may not prove to be helpful. We use Markov model based predictors, and discuss more about our predictive policies in Section 7.5 7.2.4 Taxonomy Here we describe a taxonomy of disk spindown policies, considering both the policy ....

[Article contains additional citation context not shown here]

J. Wilkes, "Predictive Power Conservation," Hewlett Packard Laboratories Technical Report HPL-CSP-92-5, , February, 1994. 136 CHAPTER 12. CONCLUSIONS

....and seriously affect user satisfaction. Instead, the program can inform the PM to prevent the OS from shutting down any of these devices. Taskspecific power management makes predictive wake up a feasible solution to improve performance while saving power. Predictive wake up was first proposed in [16] to reduce the performance penalty during wake up; however, accurate prediction is difficult for operating systems because they do not have enough knowledge about the future behavior of applications. Inaccurate prediction may either waste energy (wake up too early) or degrade performance (wake up ....

J. Wilkes. Predictive power conservation. Technical report, Hewlett-Packard, HPL-CSP-92-5, 1992.

....has to be presented to the end user with information about the energy quality tradeoff involved. 6. Related Work 6.1 Power Aware Systems There has been considerable work on power management for components of our digital camera. This work includes spindown policies for disks and alternatives [2, 5, 6, 7, 16, 24, 28, 44], and managing wireless communication [18, 23, 41, 39] Lorch et al. 29] present a survey of the various software techniques for energy management. Our camera should be able to exploit power management techniques for the components upon which it depends. Unfortunately, our usage model differs in ....

J. Wilkes. Predictive Power Conservation. Technical Report HPL-CSP-92-5, Hewlett-Packard Labs, February 1992.

....in speech recognition. From a broader perspective, there has been a substantial body of work on hardware power management for mobile computers. This includes efforts that span multiple components [10, 16] as well as those focusing on a single component such as the network [11, 20] disk [4, 5, 6, 12, 25], and CPU [14, 24] As shown in this paper, such efforts aimed at hardware components are complementary to reducing energy usage through application driven fidelity reduction. 7 Conclusion Relentless pressure to make mobile computers lighter and more compact places severe restrictions on battery ....

Wilkes, J. Predictive power conservation. Technical Report HPL-CSP-92-5, Hewlett-Packard Laboratories, February 1992.

....energy consumption. Increasing the size of the cache beyond 1 Mbyte does not help performance. This supports the findings of Douglis [4] 4. 2 Adaptive Spin down Policy As a follow up to their previous work, Douglis et al. in [6] look at an adaptive spin down policy first suggested by Wilkes [17]. They separate spin down delays into two categories unacceptable and acceptable. Their argument is that a user will not mind waiting a few seconds for the disk to spin up if s he has been idle for 30 minutes, but will mind waiting if s he has only been idle for a few seconds. Unacceptable ....

J. Wilkes, Predictive Power Conservation, Technical Report HPL-CSP-92-5, February 1992.

....the best fixed time out value computed in retrospect. Douglis et al. 4] showed that the disk sub system on portable computers consumes a major portion of the available energy (Greenawalt [7] states 30 or more) It is well known that spinning the disk down when it is not in use can save energy [4, 5, 13, 20]. 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 deciding when to spin down the disk is needed to maximize the energy savings. Current mobile ....

J. Wilkes, "Predictive power conservation," Tech. Rep. HPL-CSP-92-5, Hewlett-Packard Laboratories, Feb. 1992. A Theoretical Justification for Harmonic Idle Times We now verify that the harmonic idle times of order t have total length equal to t.

....would be effective to use a weighted average of a few previous interarrival times to decide when to spin down the disk on a mobile computer. He noted as well that if inactive intervals were of roughly fixed duration, the disk could be spun up in advance of the expected time of the next operation [Wilkes 1992]. If access patterns are not so consistent, however, these techniques may not prove to be helpful. Golding, et al. studied idle time detection and prediction in a more general framework [Golding et al. 1995] They considered a number of prediction methods, including arithmetic and geometric ....

John Wilkes. Predictive power conservation. Technical Report HPL-CSP-92-5, HewlettPackard Laboratories, February 1992.

....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 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 deciding when to spin down the disk is needed to maximize the energy savings. Current mobile ....

John Wilkes, "Predictive power conservation", Tech. Rep. HPL-CSP-92-5, Hewlett-Packard Laboratories, Feb. 1992.

No context found.

John Wilkes. Predictive Power Conservation. Technical Report HPL-CSP-92-5, HewlettPackard, May 1992.

No context found.

John Wilkes. Predictive Power Conservation. Technical Report HPL-CSP-92-5, HewlettPackard, May 1992.

No context found.

J. Wilkes. Predictive power conservation. Technical Report HPL-CSP-92-5, Hewlett-Packard Laboratories, Feb. 1992.

No context found.

J. Wilkes, Predictive Power Conservation, technical report HPL-CSP-92-5, Hewlett-Packard, Palo Alto, Calif., 1992.

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC