Minseok Song

Exploiting flash memory for reducing disk power consumption in portable media players

Portable media players are increasingly using hard disk drives (HDD) to meet their storage requirements, but HDDs consume a significant amount of energy. Therefore video frames are prefetched into dynamic random access memory (DRAM) to allow the disk to go into low-power mode; but most mobile systems have limited DRAM, so little energy is actually saved in this way. We propose two new energy saving schemes: one improves the utilization of DRAM in storing prefetched frames, and the other extends this approach by making use of auxiliary flash memory. Simulations show that deploying a reasonable amount of auxiliary flash reduces disk energy consumption by up to 86% compared with conventional prefetching schemes.

Aggressive dynamic voltage scaling for energy-aware video playback based on decoding time estimation

An effective way for reducing CPU power consumption is to reduce its operating frequency. But this slows down program execution, which may violate the real-time requirements of video playback. What is worse, it is difficult to predict future decoding times, and unduly conservative policies may miss viable opportunities to reduce the CPU frequency. The effectiveness of such power-saving techniques is thus dependent on an ability to estimate future demands on the CPU. We present the design, implementation and evaluation of a dynamic voltage scaling (DVS) scheme for portable media players. We measured decoding times on real videos and extracted a precise relationship between frame size and decoding time using logarithmic regression. Based on this model, we propose a frequency selection algorithm which accepts some deadline misses, and selects the frequencies required to achieve a specified deadline miss ratio. We implemented this scheme in MPlayer running on the Linux 2.6. Experimental results show that its system-wide energy consumption is up to 17% less than conventional DVS schemes and up to 24% less than non-DVS schemes.

Scheduling a Video Transcoding Server to Save Energy

Recent popular streaming services such as TV Everywhere, N-Screen, and dynamic adaptive streaming over HTTP (DASH) need to deliver content to the wide range of devices, requiring video content to be transcoded into different versions. Transcoding tasks require a lot of computation, and each task typically has its own real-time constraint. These make it difficult to manage transcoding, but the more efficient use of energy in servers is an imperative. We characterize transcoding workloads in terms of deadlines and computation times, and propose a new dynamic voltage and frequency scaling (DVFS) scheme that allocates a frequency and a workload to each CPU with the aim of minimizing power consumption while meeting all transcoding deadlines. This scheme has been simulated, and also implemented in a Linux transcoding server, in which a frontend node distributes transcoding requests to heterogeneous backend nodes. This required a new protocol for communication between nodes, a DVFS management scheme to reduce power consumption and thread management and scheduling schemes which ensure that transcoding deadlines are met. Power measurements show that this approach can reduce system-wide energy consumption by 17% to 31%, compared with the Linux Ondemand governor.

Energy-aware data prefetching for multi-speed disks in video servers

Energy consumption is an important issue in data centers, and disks use a significant proportion of the total energy. A promising approach to reducing disk energy consumption is to use multi-speed disks with lower rotational speeds. But changing the speed requires a significant time, which reduces the number of viable opportunities to decrease the disk speed. We propose a new energy-aware data prefetching scheme for multi-speed disks in video servers. We start by examining the power and performance characteristics of multi-speed disks. We then analyze the disk bandwidth and buffer requirements needed for jitter-free speed transitions and propose a new data prefetching scheme to extend the time that disks stay at low speed. The effectiveness of the proposed scheme is evaluated through simulations.

Adaptive disk power management for portable media players

To support the large storage requirements, consumer electronics for video playback are increasingly being equipped with hard disk drives (HDD) that consume a significant amount of energy. A video player may prefetch many frames to give disk an opportunity to go to standby mode, but this may cause playback to be distorted or stopped if timely power mode transitions are not incorporated. We present the design, implementation and evaluation of a data prefetching scheme for energy-aware video data retrieval for portable media players (PMP). We formulate the problem when the prefetching is used for variable-bit-rate (VBR) streams to reduce disk energy consumption and then propose a new energy-aware data retrieval scheme that prefetches video data in a just-in-time way so as to increase the period in which disk stays in standby mode while guaranteeing the real-time service. We implemented our scheme in the legacy video player called Mplayer that is typically used for Linux-based consumer devices. Experimental results show that it saves energy as much as 51% compared with conventional schemes.

Saving Power in Video Playback on OLED Displays by Acceptable Changes to Perceived Brightness

Displays based on organic light-emitting diodes (OLEDs) are now widely used in mobile devices, in which they are major power consumers. The power drawn by an OLED display increases nonlinearly with sub-pixel intensities-thus reducing brightness saves appreciable power, but can displease users. This paper examines this tradeoff, and proposes a color blending scheme in which each frame is darkened in a way that reduces power consumption significantly while limiting the visual impact. The target lightness of frames is determined, in the LAB color space, from average intensities of the original frames; and these average values are in turn obtained by adaptive sampling, so as to reduce the computational overhead. An Android smartphone uses 12%-36% less power when running this scheme, compared to standard video playback, while a user evaluation suggested that reducing brightness to a certain extent can be largely unnoticed, or readily tolerated.

계층적 비디오 코딩의 품질확장성을 활용한 전력 관리 기법

미디어 플레이어에서의 디코딩 과정은 많은 연산을 필요로 하며, CPU로부터 높은 소비전력을 초래한다. 디코딩 연산을 줄이는 것은 CPU 소비 전력을 감소시킬 수 있지만 사용자로부터 비디오 품질을 저하시키게 된다. 본 논문에서는 H.264의 품질 확장성을 이용하여 새로운 CPU 전력 관리 기법을 제안한다. 첫째, VQM(Video Quality Metric)을 사용하여 계층적 비디오 코딩의 서로 다른 양자화 인자를 고려한 새로운 비디오 품질 모델을 제안한다. 그리고 이전 디코딩 시간과 프레임 크기를 선택적으로 융합한 디코딩 시간 예측기법에 기반한 새로운 동적 전압 기법을 제안한다. 최신 스마트폰에서 구현하였고, 사용자 테스트를 수행하였다. 제안한 기법을 실제 측정에 적용하였을 때 리눅스 동적 전압 및 주파수 조절(DVFS) 거버너에 비해 34%의 에너지 감소를 보였고 사용자 테스트를 통해 실험 영상의 품질 하락을 사용자는 인지하지 못하거나 용인될 수 있음을 확인하였다.

Balancing energy use against video quality in mobile devices

It is now common to access mobile video services through various Internet-connectable devices such as game stations, smartphones and tablets. This paper proposes a new energy management scheme for mobile devices by making use of the temporal scalability available in scalable video coding (SVC) techniques. The proposed scheme determines a quality index for each video segment, based on the pixel-wise difference between the luminance planes of consecutive frames at each frame-rate, and predicts the computation needed to decode each frame using data relating frame sizes to previous decoding times. A new dynamic voltage and frequency scaling (DVFS) scheme is then invoked to achieve the specified quality index. The proposed scheme was implemented on a smartphone and a user test was conducted to assess the acceptability of different degree of frame-rate reduction. Real measurements showed that the proposed scheme uses between 5% and 41% less energy than the Linux DVFS governor used in many smartphones, while maintaining acceptable video quality.

Saving Energy in Video Servers by the Use of Multispeed Disks

Energy consumption is an important issue in data centers, and disks use a significant proportion of the total energy. A promising approach to reducing disk energy consumption is to use multispeed disks with lower rotational speeds, and allowing disks to run slowly when workloads are light can reduce their large contribution to the power used by video servers. We formulate the prerequisites for speed reductions, and derive an energy model constrained by retrieval period. We then analyze the relationship between retrieval period, buffer size, and disk speed, and examine the effect of buffer allocation on energy consumption. We then propose a new video data retrieval scheme in which the retrieval period and the speed of each disk is dynamically changed to adjust disk bandwidth utilization, with the aim of allowing disks to run at lower speeds while guaranteeing jitter-free speed transitions. Experimental results show that our scheme achieves appreciable energy savings under all workloads. They also reveal that increasing the number of available speeds reduces energy consumption but the benefits gradually tail off.

Replication and retrieval strategies for resource-effective admission control in multi-resolution video servers

Video-on-demand (VOD) service requires balanced use of system resources, such as disk bandwidth and buffer, to accommodate more clients. The data retrieval size and data rates of video streams directly affect the utilization of these resources. Given the data rates which vary widely in multi-resolution video servers, we need to determine the appropriate data retrieval size to balance the buffer with the disk bandwidth. Otherwise, the server may be unable to admit new clients even though one of the resources is available for use. To address this problem, we propose the following new schemes that work together: (1) A replication scheme called Splitting Striping units by Replication (SSR). To increase the number of admitted clients, SSR defines two sizes of striping unit, which allow data to be stored on the primary and backup copies in different ways. (2) A retrieval scheduling method which combines the merits of existing SCAN and grouped sweeping scheme (GSS) algorithms to balance the buffer and disk bandwidth usage. (3) Admission control algorithms which decide whether to read data from the primary or the backup copy. The effectiveness of the proposed schemes is demonstrated through simulations. Results show that our schemes are able to cope with various workloads efficiently and thus enable the server to admit a much larger number of clients.