Pre-fetching and Caching in Catch-Up TV Network

Abstract

Nowadays, huge amounts of data are transferred over the Internet. Video content is a vast majority of the data traffic. A part of this data is wasted due to many reasons. Transmitting large amounts of data across a network leads to network congestion and start-up delay of the video playback, which is one of the reasons of data loss. Another reason is the negative effect of impatient users or so-called zappers. This effect is expressed by the unexpected change in user preference during the streaming process of the requested content. In that case, the downloaded data are discarded which can be considered as a waste of network and system resources.

One of the solutions to this problem is to develop a system for prefetching and caching, which is used to prepare the expected contents to be ready for any possible request by the users. The improvement in the system is approached by eliminating excess data transfer, and by caching needed parts of the contents. Thus reducing the load on the network and saving the resources.

The aim of this study is to investigate user behavior and define user groups as Loyals or Zappers on a scale of a grading system. This is performed by analyzing collected data of user requests and content details to find a way to adjust prefetching and caching system settings based on several factors; user behavior, session length, content length, and content popularity. An individual calculation for each of these factors is done to get the specified results which are shown in graphs to have an overview of the analyzed data, and to extract useful information to reach the goal of this work. All of the studied subjects are contributed to produce an enhanced model of the prefetching and caching system.

After demonstrating the results, some variable values can be evaluated from the calculations. These values vary depending on the processed data. That would affect the accuracy of the outcomes.

Popular Abstract

Internet TV has been more available and widely popular recently. A fast Internet broadband connection, which can deliver high-quality videos, is one of the reasons for the flourishing of the Internet Protocol TV and Video-on-Demand service. Another reason is the possibility to watch TV contents from around the world over the Internet, at any place in the world and at any time. Catch-up TV is a service which permits the viewers to access TV programs and video contents beyond the scheduled original broadcasting time.

Prefetching and caching techniques have been used to reduce the start-up delay and the latency in viewing online video streams. Prefetching is predicting the possible videos that the user may be interested to watch next, and preload a part of them in the cache, before they have been requested by the user. The cache is a fast memory located at the server, and also can be located at the end-user's terminal. A part of the frequently requested video, or the predicted video, can be stored in the cache. The benefit from video caching is to decrease the delay of retrieving the videos from the main storage of the server, or from its original source, and to make the videos ready to be viewed by the user in a fast way.

The first issue that faces the Catch-up TV service is the unused data transfer. Preloading a lot of videos that may not be watched by the user, is a waste of the system resources. As a result, the bandwidth will be exhausted, and the cache will be filled with unnecessary data. The second issue is the increasing usage of the Internet TV service that can cause server overload with many requests from the users, and leads to network congestion, data loss, and degradation of video streaming quality.

Analyzing the usage of the service has an emerging importance to enhance the service by allocating system and network resources based on the behavior of the users and the requirements of the service provider. The users do not have the same interest in the available video contents of the service, many users start watching a video, then leave it right in the beginning, and change to another video, other users continue watching the requested video to the end. The impatient users, who constantly change their watching preference at the beginning of the video stream, have a great impact on the network, that is a lot of data will be transmitted over the network with no use. The unused data forms an excessive load on the server and the network. The varied user activity should be taken into the account when deciding system settings, and when distributing system and network resources.

The aim of this project is to analyze user behavior in catch-up TV network in order to identify user viewing habits to provide the required resources to each user. We strive to find an adaptive caching and prefetching algorithm that can be implemented to decrease the amount of cached data in the server and the user cache, and also to decrease the transmitted data over the network. Studying user behavior and trying to use new methods of prefetching and caching techniques, achieves twofold benefits, both for the service provider by saving system and network resources, and also for the users by improving video streaming quality and decreasing time delay.

There are many factors that can be used to achieve the goal of this research. Those factors are related to both user behavior and content properties. The factors related to the user concern the preference of the user based on the viewing history of the user, and also about how long the user watches the contents in average. On the other hand, the factors related to the contents concern its popularity, length, and type.

We have studied each factor separately to one value denotes the result from each studied section that summarizes the user activity or the content properties. The resulted value from each section can be combined together to get an ultimate result to decide how much of the data is needed to be cached for each predicted content, and how much data is needed to be transferred to each user. All those factors can be used cooperatively to determine the right amount of the data to be cached in the server cache and in the user cache. The cache of the server is more related to the common contents, which are viewed by many users. Furthermore, the cache of the user is related to the activity of each specified user.

Since different users have different preferences, which can change all the time, even for the same user, the data analysis should be done in real time in the server to achieve an actual outcome that can be used to modify the setting of the system to get a better result.

• author: Jaddoa, Wael and Abbas, Housam
• supervisor: Jens A Andersson; Stefan Höst; Maria Kihl
• organization: Department of Electrical and Information Technology
• course: EITM02 20161
• year: 2017
• type: H2 - Master's Degree (Two Years)
• subject: Technology and Engineering
• keywords: Prefetching, Caching, Cache, Buffering, Catch-up TV, VOD, SVOD, OTT, IPTV, Loyal, Zapper
• report number: LU/LTH-EIT 2017-594
• language: English
• id: 8919690
• date added to LUP: 2017-06-29 14:06:38
• date last changed: 2017-06-29 14:06:38