Data-Prefetching Scheme Based on Playback Delay and Positioning Satisfaction in Peer-To-Peer Video-On-Demand System

Lei Wang, Xiaorui Li, Yaqiu Liu, Guan Gui
2018 Sensors  
As one of the most important applications in peer-to-peer (P2P) networks, the video-on-demand (VoD) system freely supports video cassette recorder (VCR) operation for users. However, the users may experience significant playback delay after frequent VCR operations in the VoD system, which will affect the quality of experience (QoE) of the users. Hence, selecting an appropriate data-prefetching strategy to support better VCR operation is an important approach to improve the QoE. This paper
more » ... es a data-prefetching strategy (DSA) to determine the most suitable anchor interval by considering the playback delay and positioning satisfaction. According to the DSA, we use the multiple-attribute decision-making (MADM) theory to model the selection of intervals of prefetching data blocks (i.e., anchor interval) and the technique for ordering preference by similarity to an ideal solution (TOPSIS) algorithm to solve the MADM. The simulation results show that the DSA strategy obtains higher positioning satisfaction than the existing schemes, which is approximately 60% higher than the anchor points, popular parts of video, and user interests (API)-based method. Moreover, with the increase in network bandwidth, the DSA strategy can minimize the playback delay after VCR operation using relative few extra bandwidths. Sensors 2018, 18, 816 2 of 13 selection algorithm to reduce the playback delay after the VCR operation. A data-prefetching strategy was proposed, i.e., if residual download bandwidth remains, the peer can prefetch a part of non-urgent data after the urgent data has been downloaded. The urgent data can ensure the video is continuously played in the next several minutes, and the non-urgent data is used to reduce the playback delay after the users perform VCR operations. In this strategy, the peer uses the residual bandwidth to download the data ahead of the current playback point, which can reduce the playback latency and effectively improve the viewing experience of users [4] . In [5] , Yu et al. analyzed the VCR operation behaviors of users in large-scale VoD systems, and proposed a strategy to reduce the playback delay after the VCR operation by prefetching videos with high play counts. This strategy assumes that all users in the system have identical preferences, so they attempt to predict the video that is most likely to be played in the next few minutes by analyzing a large volume of user viewing logs. However, the assumption has drawbacks. For example, the viewing preferences of users vary for different times. Hence, inaccurate predictions will cause a low bit rate of prefetched data and waste bandwidth resources. In [6], the authors proposed an anchor-prefetching strategy. The anchor is distributed throughout the video file with a fixed interval (5 min), and each anchor consists of 10 continuous seconds. This traditional anchor-prefetching strategy can effectively reduce the playback delay after the VCR operation and avoid the problem of bandwidth waste caused by inaccurate prediction. The authors of [7] developed a prefetching strategy of API by prefetching both anchor points and popular segments of the video. This approach is complementary to the anchor-prefetching strategy, which can reduce the playback delay and improve the user satisfaction. However, to implement this strategy, it is necessary to detect the video segments that are currently played by each online P2P node, which will consume many extra bandwidth resources. In [8] , the authors measured and analyzed the rules of VCR operation when the user watches the video. The forward-and backward-jump lengths have been modeled with a Weibull distribution. The forward and backward distances have different parameter values of Weibull distribution. In addition, for different lengths of video files, the Weibull distribution parameter values also vary. This is the theoretical basis of our study on the data-prefetching strategy. A new strategy to determine the most reasonable anchor interval is proposed in this paper by considering two factors of playback delay and positioning satisfaction. How to minimize the playback delay after the VCR operation is one of the most serious challenges in the design of P2P-based VoD systems, as mentioned above. And as mentioned in [6], in a VoD system with the anchor-prefetching strategy, the playback point is adjusted to the beginning of the closest anchor after the VCR operation. The disparity between the actual playback point and the designated point by the VCR operation may reduce the user's viewing experience. Therefore, we introduce the concept of positioning satisfaction to reflect the user's appraisal of the anchor-based VoD system caused by the disparity. Then, the user's viewing experience is studied by considering the factors of playback delay and positioning satisfaction. The playback delay is a cost factor, and a lower value indicates better viewing experience of the users. The positioning satisfaction is a benefit factor, and a higher value indicates better viewing experience of the users. The problem of selecting the most reasonable anchor interval was modeled using the multiple-attribute decision-making (MADM) theory [9] . Based on the factors of playback delay and positioning satisfaction, the MADM model is established. By solving this model using the technique for ordering preference by similarity to an ideal solution (TOPSIS) method, the optimal anchor interval is selected from 20 optional anchor interval schemes. The remainder of the paper is structured as follows. The model of the P2P-based VoD system is described in Section 2. The problems studied in this paper are described in Section 3. The proposed optimal selection strategy of anchor interval is presented in Section 4. The performance evaluation is presented in Section 5. Section 6 concludes this paper.
doi:10.3390/s18030816 pmid:29518014 pmcid:PMC5876552 fatcat:i6d2jviyuncnpgmluro6dv2pp4