Video Delivery Networks Research Articles

Optimal QoE-Fair Resource Allocation in Multipath Video Delivery Network

A steadily increasing number of users consume videos over the Internet. In current video platforms, players run a control algorithm that dynamically chooses the video bitrate to match the time-varying network bandwidth. Such an algorithm strives to improve the quality individually perceived by users. Consequently, this control architecture leads, in the optimal case, to maximize the average quality perceived collectively by all users rather than to a quality-fair distribution of resources, possibly leading to user abandonment for those users receiving a lower quality. Therefore, we argue that well-designed video delivery networks should gracefully degrade the perceived quality equally for all users when resources become scarce. In this paper, we propose the Multi-Commodity Flow Problem (MCFP) optimization framework to address the issue of designing a QoE-fair optimal allocation strategy. We show how to make the resulting problem tractable for video platforms serving massive audiences. The performance of the proposed optimal fair resource allocation strategy is tested through realistic simulations involving thousands of concurrent users on two real networks by varying both the total load on the network and the system parameters.

Content to cash: Understanding and improving crowdsourced live video broadcasting services with monetary donations

Crowdsourced live video broadcasting (livecast) services such as Twitch and Douyu have become increasingly popular in recent years. In such a service, how to allocate limited service capacities, including video transcoding and delivery capacities among numerous channels, is a critical problem. Previous studies allocate capacities based on popularity. In this paper, we analyze Douyu, a leading crowdsourced livecast website in China, with a measurement approach. We find that Douyu is deployed upon a video delivery network (VDN), and it prioritizes popular channels when allocating service capacities; we also find that viewers’ willingness to donate monetary gifts in a channel is closely related to their streaming experiences, which are decided by service capacities allocated in the channel. On the other hand, a livecast channel’s profitability is only moderately correlated to its popularity. In other words, there exists a mismatch between the popularity-based service strategies and Douyu’s business model. Motivated by our analysis, we propose that channels’ profitability as well as popularity should be considered in capacity allocating. We present proactive and reactive algorithms, which balance viewers’ satisfaction with system’s monetary profit, for allocating transcoding capacity among livecast channels. We also propose a practical strategy for VDN edge nodes to select channels to replicate, by taking channels’ popularity, profitability, and bandwidth consumptions into consideration. Experiments driven by real-world measurement data show that our proposed solutions can effectively improve the overall benefits for a crowdsourced livecast system and individual VDN edge nodes, and avoid adjusting channels’ transcoding schemes too often during livecast sessions.

A Multi-Commodity Flow Problem for Fair Resource Allocation in Multi-Path Video Delivery Networks

Video streaming services employ the Internet to distribute content to an ever-increasing number of concurrent viewers. The delivery architecture employed by leading video platforms requires players to run a control algorithm dynamically choosing the video bitrate to match the time-varying network bandwidth and avoid playback interruptions due to buffer underruns. Such an algorithm is generally designed to (selfishly) improve the quality individually perceived by users. Consequently, this control architecture leads, in the optimal case, to maximize the average quality perceived collectively by all users and not to a distribution of resources that is fair in terms of user perceived quality. We argue that video service providers should manage their delivery network to address fairness issues to gracefully degrade the perceived quality equally for all users when resources become scarce. Even though the general problem of providing a fair level of perceived quality does not scale with the cumbersome number of concurrent users, this paper shows that the Multi-Commodity Flow Problem (MCFP) optimization framework is a proper and efficient tool to address this open issue. First, we show how to cast the resource allocation problem to an MCFP and then we propose a strategy to make the resulting problem tractable for video distribution platforms serving massive audiences. The performance of the proposed optimal fair resource allocation strategy is assessed using realistic simulations involving thousands of concurrent video sessions on a real network topology by varying both the total load on the network and key system parameters.

Video Content Delivery Schemes: Approaches and Directions

This paper addresses the main two types of video delivery networks deployed into digital television services throughout the years. Underlying and Overlay networks are described in order to introduce how the physical infrastructure and network architecture impacted on the modeling of the latest video content delivery schemes. The convergence of different video delivery networks, in a technical sense, is a real thing and may disrupt the broadcasting industry in the long-term.

Cost Optimization for the Coupled Video Delivery Networks

Video delivery networks (VDNs) are one of the key enablers for alleviating contradiction between demands increasing for online video and response time of users accessing video content. However, existed works mainly adopt either proactive or reactive delivery scheme, which leads to the problems of backbone traffic overloaded and delays increased. In these works, the impact of performance indicators, such as bandwidth, delay, and personalized demand on delivery cost are not comprehensively taken into account. To overcome these problems, we propose a cost optimization for coupled video delivery model (CO-CVDM), which combines coupled proactive delivery and coupled reactive delivery method to minimize delivery cost in terms of bandwidth consumption, delay performance, personalized demand and construct a multicast delivery tree for the video delivery. The coupled delivery cost minimization problem under constraint conditions is formulated. Since this problem is NP-hard and is prohibitively difficult to solve, we develop a heuristic multicast delivery tree (HMDT) algorithm to provide more optimized approximate cost. Based on this, we extend CO-CVDM to distribute video in a dynamic scenario. Specifically, with dynamic change of user's requirement for video, the process of video delivery is truncated into a sequence of static graphs and the change of interest attenuation is modeled with human forgetting curve. Finally, the numerical simulations are conducted to validate the advantage of CO-CVDM in terms of optimization delivery cost, and the delivery cost is inversely proportional to the attenuation of user interest.

Detection of Delivery Servers in Online Video Service using Long Short Term Memory Network

Nowadays large scale video delivery networks have been widely deployed. Information about video delivery servers is crucial for a series of network management tasks. However, it is a challenge to identify the delivery servers in the ever-changing content delivery networks. Moreover, it is difficult to find general and flexible features to identify video delivery servers from various video service providers. In this paper, we mainly come up with a novel method based on LSTM (Long Short Term Memory Network) to detect video delivery servers. Experimental results prove that our approach works pretty well by comparing it with the other two conventional machine learning methods. What’s more, our approach can work more flexibly without cumbersome feature extraction relative to conventional machine learning methods.

Session-Based Cloud Video Delivery Networks in Mobile Internet

Cloud video delivery networks (C-VDNs) have been receiving considerable attention recently. Different from traditional video delivery networks, C-VDNs can provide scalable resources without the need to invest on infrastructure deployment. However, current C-VDN strategies cannot be readily applied to mobile Internet because of the dynamic characteristics of mobile users and videos. Moreover, dealing with the surrogate site selection and delivery path building under cloud resource constraints is another challenging problem. We propose the session-based cloud video delivery network framework to solve these problems. This framework considers the mobility of mobile users and the changing of videos by employing graph extension and session construction, respectively. Furthermore, we formulate the coordination of surrogate site selection and delivery route building as a mixed integer program. We relax the integer constraints and design a heuristic algorithm to obtain the optimal solution because this is a nondeterministic-polynomial time problem. Simulation experiments show that our proposed algorithm is efficient in terms of both performance and rental cost.

Video delivery networks: Challenges, solutions and future directions

Internet video ecosystems are faced with the increasing requirements in versatile applications, ubiquitous consumption and freedom of creation and sharing, in which the user experience for high-quality services has become more and more important. Internet is also under tremendous pressure due to the exponential growth in video consumption. Video providers have been using content delivery networks (CDNs) to deliver high-quality video services. However, the new features in video generation and consumption require CDN to address the scalability, quality of service and flexibility challenges. As a result, we need to rethink future CDN for sustainable video delivery. To this end, we give an overview for the Internet video ecosystem evolution. We survey the existing video delivery solutions from the perspective of economic relationships, algorithms, mechanisms and architectures. At the end of the article, we propose a data-driven information plane for video delivery network as the future direction and discuss two case studies to demonstrate its necessity.

A model for evaluating sharing policies for network-assisted HTTP adaptive streaming

HTTP adaptive streaming (HAS) has become the dominant technology for streaming video over the Internet. It gained popularity because of its ability to adapt the video quality to the current network conditions and other appealing properties such as usage of off-the-shelf HTTP servers and easy firewall traversal. However, when multiple HAS players share a bottleneck link for streaming, the individual adaptation techniques in the players have difficulties to maintain a stable bitrate and fairly share the network resources. HAS-assisting network elements can solve these performance problems and allow execution of advanced policies for sharing the available bandwidth. Nonetheless, testing and evaluating new sharing policies is costly and time consuming. This motivated us to formulate a model that allows to differentiate between groups of users depending on the type of user or device, and that can describe the mean bitrate of the video streams and how often this bitrate is expected to change during playout. To show how our model can be used, we demonstrate two applications of our model. Furthermore, we validate the model based results against results obtained using our streaming testbed and proxy server based HAS-assistant. The results show that our model is highly accurate for both the mean bitrate and the number of changes in video bitrate. As such, our model is a useful tool for network administrators and internet service providers for evaluating the performance of sharing policies and for managing and provisioning video delivery networks.

Practical, Real-time Centralized Control for CDN-based Live Video Delivery

Live video delivery is expected to reach a peak of 50 Tbps this year. This surging popularity is fundamentally changing the Internet video delivery landscape. CDNs must meet users' demands for fast join times, high bitrates, and low buffering ratios, while minimizing their own cost of delivery and responding to issues in real-time. Wide-area latency, loss, and failures, as well as varied workloads ("mega-events" to long-tail), make meeting these demands challenging. An analysis of video sessions concluded that a centralized controller could improve user experience, but CDN systems have shied away from such designs due to the difficulty of quickly handling failures, a requirement of both operators and users. We introduce VDN, a practical approach to a video delivery network that uses a centralized algorithm for live video optimization. VDN provides CDN operators with real-time, fine-grained control. It does this in spite of challenges resulting from the wide-area (e.g., state inconsistency, partitions, failures) by using a hybrid centralized+distributed control plane, increasing average bitrate by 1.7x and decreasing cost by 2x in different scenarios.

Joint assessment of Network- and Perceived-QoS in video delivery networks

Given the increasing number of IP streaming video customers, service providers are seeking an efficient way to monitor in real time the offered quality of service, as perceived by each end user. Since real-time video quality assessment via image processing algorithms is quite bandwidth- and processing-power-demanding, a feasible alternative could be to monitor the Network-level Quality of Service (NQoS) and associate it with the Perceived QoS (PQoS). This article presents a network-agnostic framework for the joint assessment of N- and P-QoS, with the aim of correlating these two parameters for a specific network and service configuration. This framework/architecture is implemented with open source software tools and is being demonstrated in an actual WiMAX streaming video distribution platform.

Automatic quality of experience measuring on video delivering networks

This article describes a full video delivery network monitoring suite. Our monitoring tool offers a new view of a video delivery network, based on the quality as perceived by final users (what is nowadays called Quality of Experience, in short QoE). We measure the perceived quality at the client side by means of the recently proposed PSQA technology, by studying the video flows at the frame level. The developed monitoring suite is a completely free-software application, based on well-known technologies such as Simple Network Management Protocol or Round Robin Databases, which can be executed in various operating systems. In this short article we explain the tool implementation and we present some of the measurements performed with it.

The Video-On-Demand Trial in Taiwan: A Partially DAVIC Compliant System

This paper reviews the Video-On-Demand technical trial conducted by Computer & Communication Research Laboratories in Taiwan. This trial is one of the key realization of Taiwan‘s National Information Infrastructure. The primary goal is to provide a testbed and platform for new product development, interoperability test, and interactive multimedia applications. It also explores the technical difficulties from which experiences can be gained for future large scale deployment. Open system and end-to-end system are the most important requirements of the VOD system in this trial. Open system allows a variety of video servers and set-top boxes to be integrated into the system while end-to-end system requires system integration know-how. An overview of the Video-On-Demand system is provided in this paper. The architecture of the video server, delivery network and set-top box will be examined. The system integration issues including content preparation will also be discussed.