Yishuai Chen

Measurement and Modeling of Video Watching Time in a Large-Scale Internet Video-on-Demand System

Video watching time is a crucial measure for studying user watching behavior in online Internet video-on-demand (VoD) systems. It is important for system planning, user engagement understanding, and system quality evaluation. However, due to the limited access of user data in large-scale streaming systems, a systematic measurement, analysis, and modeling of video watching time is still missing. In this paper, we measure PPLive, one of the most popular commercial Internet VoD systems in China, over a three week period. We collect accurate user watching data of more than 100 million streaming sessions of more than 100 thousand distinct videos. Based on the measurement data, we characterize the distribution of watching time of different types of videos and reveal a number of interesting characteristics regarding the relation between video watching time and various video-related features (including video type, duration, and popularity). We further build a suite of mathematical models for characterizing these relationships. Extensive performance evaluation shows the high accuracy of these models as compared with commonly used data-mining based models. Our measurement and modeling results bring forth important insights for simulation, design, deployment, and evaluation of Internet VoD systems.

A Measurement Study of Cache Rejection in P2P Live Streaming System

In P2P system, each peer caches the received data for accessing of other peers. P2P live streaming system audiencepsilas TV-like watching pattern brings unique characteristics for its cache management algorithm. By measuring and studying the cache rejection algorithm used in PPLive with our PPLive crawler, we found a simple but interesting cache rejection algorithm used in PPLive, i.e. the cache rejection of peers is synchronized with the chunk upload of media server on chunk offset. We modeled this algorithm as a fixed-duration virtual buffer which includes the P2P network and peerpsilas local buffer and then verified it in our trace. We finally gave some analysis and explanations for this algorithm. What we found is benefit to the future design and modeling of P2P live streaming system.

Performance modeling and evaluation of peer-to-peer live streaming systems under flash crowds

A peer-to-peer (P2P) live streaming system faces a big challenge under flash crowds. When a flash crowd occurs, the sudden arrival of numerous peers may starve the upload capacity of the system, hurt its quality of service, and even cause system collapse. This paper provides a comprehensive study on the performance of P2P live streaming systems under flash crowds. By modeling the systems using a fluid model, we study the system capacity, peer startup latency, and system recovery time of systems with and without admission control for flash crowds, respectively. Our study demonstrates that, without admission control, a P2P live streaming system has limited capacity to handle flash crowds. We quantify this capacity by the largest flash crowd (measured in shock level) that the system can handle, and further find this capacity is independent of system initial state while decreasing as departure rate of stable peer increases, in a power-law relationship. We also establish the mathematical relationship of flash crowd size to the worst-case peer startup latency and system recovery time. For a system with admission control, we prove that it can recover stability under flash crowds of any sizes. Moreover, its worst-case peer startup latency and system recovery time increase logarithmically with the flash crowd size. Based on the analytical results, we present detailed flash crowd handling strategies, which can be used to achieve satisfying peer startup performance while keeping system stability in the presence of flash crowds under different circumstances .

Measure and Model P2P Streaming System by Buffer Bitmap

The correct evaluation of P2P streaming system models needs the validation in real world system. However, there is lack of systematic and integrated measurement method for real world P2P streaming system. In this paper, we propose a P2P streaming network measurement method based on a peers buffer occupancy probability. Our method is based on the fixed duration buffer property of commercial P2P streaming systems. We prove the measured buffer occupancy probability reflects the chunk propagation process in the P2P network. We then propose a P2P streaming chunk propagation model and verify it in the commercial P2P streaming network using our measurement method. Our measurement method is useful for measuring and analyzing miscellaneous P2P streaming systems. And our model and parameter estimation are useful for existing P2P simulators to choose correct parameters and are meaningful for researchers to understand the real meaning behind the parameters.

Performance Investigation on Request Collisions in Peer-To-Peer Live Streaming Systems: Problems and Algorithms

A fundamental design issue in a peer-to-peer (P2P) live streaming system is how peers select neighbors for downloading media pieces. Inappropriate selection of neighbor peers for piece downloading can exacerbate the collisions of requests among peers and therefore degrade the system performance. In this paper, we first show how severe the request collision problem could be and also to which degree it could affect the performance of a P2P live streaming system by extensive measurements and simulations. We then propose a novel neighbor peer selection algorithm for piece downloading. In the algorithm, each peer needs to infer the severity of request collisions at each of its neighbors based on the most recent request-with-no-response history. Based on this information, each peer always sends requests to those neighbors with low collision probability at high priority. This algorithm can effectively avoid the difficulty in obtaining accurate network conditions and/or peer availability and also greatly reduce the signaling overhead for exchanging residual-capacity-related information among peers. In addition, the algorithm is easy to implement. Simulation results show that the proposed algorithm can effectively alleviate the request collision problem and improve the system performance.

A measurement study on user behavior of p2p VoD system

Nowadays, as the fast growth in population of p2p VoD system, user behaviors are playing an increasingly crucial role for the performance of system. In this paper, the user watching and sharing behaviors of p2p VoD system were studied by analyzing a large amount of real data. Then, a comparison between user behavior in p2p VoD system and other main kinds VoD systems was made. After that, a mathematic model was constructed to describe the characters based on those analyses. Finally, the correctness of conclusions was verified by an experiment on real system and an improved method was provided. Those studies were believed can possess guiding significance for analyzing system performance, traffic loads and studies on user behavior of other online applications.

Understanding and searching the online video in China

Nowadays, as the fast expansion of broadband internet access, users tend to simply view online video in web browsers, instead of download to local machine before watching. We investigate the necessity of introducing a general video search engine, especially for China internet users, and designed our indexing strategies and algorithms. We present our investigation result and our design, along with the influence of our search engine by studying user behavior of search and clicks.

On distribution of user movie watching time in a large-scale video streaming system

Video watching time is a crucial measure for studying user watching behavior in online Internet video-on-demand (VoD) systems. It is important for system planning, user engagement study, and service quality evaluation. However, due to limited access to large-scale VoD systems, there is still a lack of accurate model for characterizing the distribution of user watching time on a per video basis. In this paper, we measure PPLive, one of the most popular commercial Internet VoD systems in China, over a three week period, and characterize user watching time distributions of 1,000 most popular movies. We find that a videos watching time can be modeled by a concatenation of exponential distribution (in the first several minutes of the video) and truncated power law distribution (in the remaining time of the video), when users watch the video without interruptions. For comparison, user watching time with user interactions such as seeking and/or pause operations does not follow such a distribution. We further reveal interesting characteristics regarding the relation between videos watching time distribution and various watching/video-related features (including time-of-day, user ratings, and movie genres). Our measurement and modeling results bring forth important insights for design, deployment, and evaluation of Internet VoD systems.

Scheduling piece requests blindly and randomly for peer-to-peer live streaming

Abstract The design of piece scheduling algorithm is fundamental in mesh-based peer-to-peer (P2P) live streaming systems. This article regards the piece propagation process in P2P live streaming system as a conflict process, where peers compete to get the piece as soon as possible to meet the tight timing requirement of the live system. In this vision, the authors reduce the piece scheduling problem as a conflict resolution problem, and propose a blind random scheduling method to resolve it. The algorithm can be described as: 1) pieces are scheduled independently; 2) conflicts are resolved by the random back off algorithm. The theoretical analysis and simulation results prove that the algorithm scales well with the network size, and can obtain near-optimal in-time delivery of pieces. The algorithm is also easy to implement in the real-world system.

Can user privacy and recommendation performance be preserved simultaneously

Abstract In online systems of videos, music or books, users’ behaviors are disclosed to the recommender systems to learn their interests. Such a disclosure raises a serious concern in the public for the leak of users’ privacy. Meanwhile, some algorithms are proposed to obfuscate users’ historical behavior records to protect users’ privacy, at the cost of degradation of recommendation accuracy. It is a common belief that such tradeoff is inevitable. In this paper, however, we break this pessimistic belief based on the fact that peoples interests are not necessarily limited to items which are geared to a certain gender, age, or profession. Based on this idea, we propose a recommendation-friendly privacy-preserving framework by introducing a privacy-preserving module between a recommender system and user side. For instance, to obfuscate a female users gender information, the privacy-preserving module adds a set of extra factitious ratings of movies not watched by the given user. These added movies are selected to be those mostly watched by male viewers but interesting the given female user. Extensive experiments show that our algorithm obfuscates users’ privacy information, e.g., gender, efficiently, but also maintains or even improves recommendation accuracy.