Interactive Video Delivery Services

Interactive Video Delivery Services Video-On-Demand Interactive Services Interactive video delivery services are a fundamental change in the TV interface  paradigm. They shift the delivery paradigm from carrying many simultaneous parallel  streams (channels) to one that carries concurrent accesses through separate channels into a  database. Traditionally, in a broadcast TV system, many stations broadcast their programs  simultaneously and the user selects a specific channel to view. As a result, a user is  restricted to a chronology of parallel and competing programming whereas, an interactive  system makes all programming available to its users without this restriction. There is no  temporal restriction. All programming becomes available any time to the user. Types of Interactive Services Based on the amount of interactivity allowed (adapted from [4]), interactive services can be classified into several categories. The user is a passive participant and has no control over the session in broadcast (No-VOD) services that are similar to broadcast TV. The user signs up and pays for specific programming, similar to existing CATV PPV services in pay-per-view (PPV) services. The users are grouped based on a threshold of interest in quasi video-on-demand (Q-VOD) services. By switching to a different group, users can perform rudimentary temporal control activities. The functions like forward and reverse are simulated by transitions in discrete time intervals (on the order of 5 minutes) in near video-on-demand (N-VOD) services. The multiple channels with the same programming skewed in time [5, 15] can provide this capability. The user has complete control over the session presentation in true video-on-demand (T-VOD) services. The user has full-function VCR (virtual VCR) capabilities including forward and reverse play, freeze, and random positioning. For T-VOD, only a single channel is necessary; multiple channels become redundant. Technological Inhibitors There are other inhibiting issues to the ubiquitous deployment of interactive multimedia applications than just technological issues. In the digital environment, information is readily copied, reproduced, and altered, jeopardizing the established markets of the information providers. To convince an information provider to accept an all-digital system, certain incentives like mechanisms like encryption to protect intellectual property rights – that will maintain their data and thus help them stay in business are needed. (The Internet does not copy data, people copy data.) System Components for Video-on-Demand 5A detailed analysis of these issues is beyond the scope of this paper. An interesting survey of the  intellectual property rights problem has been provided by Samuelson [14]. Hundreds (if not thousands) of users with different viewing preferences will access a VOD system simultaneously. The quality of each session must remain within specified bounds to achieve customer satisfaction. This ensures the quality of the system. We will survey the individual technologies in the context of an end-to-end architecture for a VOD system. A typical VOD scenario contains a local database and server connected to user homes  via a communications network. The user home consists of a network interface coupled to a  display [4]. The user interacts with the system via a mouse or a computer keyboard. Fig. 2 illustrates this architecture. user interface and display high-speed backbone local database local server home viewer network interface multimedia archive and distributor multimedia archive Figure 2: A Simple VOD Architecture Management of System Resources in VOD We identified some of the technical problems in designing a VOD system in the previous sections. A VOD system is required to support a large customer population and many movie titles. Most existing prototypes are constricted to laboratory or office environments and support at most a few hundred users and up to a hundred movies. Large scale commercial systems  should need to more closely match the per-user resource requirements and usage patterns to  achieve economic feasibility. In this section, we look over some of these problems and discuss  existing research in this area. Resource Reservation One of the fundamental problems in developing a VOD system is one of storage and network I/O  bandwidth management. The VOD system possesses a finite amount of resources measured in  terms of storage I/O and communication bandwidths. As various customers compete for the same  system resources, efficient schemes that ensure fairness of allocation have to be designed. The service provider wants to generate the maximum revenue from the offered services. A  balance between these two often opposing requirements is necessary to tap the potential  benefits of the system. The first step to solve this problem is the development of an  accurate system model. We use the model proposed in Fig 2 as the basis for the remainder  of this discussion. The end-to-end VOD system comprises of three basic components; the storage server,  the network, and the user interface. The metadata server provides an additional level of  complexity to the system model. The time dependency of continuous media requires the  VOD system to ensure that the data transmission mechanism can provide for strict deadlines.   If these deadlines are missed, it is possible for the quality of the session to degrade. To ensure customer satisfaction, resources should be reserved along the entire data path of a connection on a per-session basis. The complexity of the resource reservation mechanism depends on the  application under consideration. Interactive services need the resource reservation to be made per-session along the entire data path, including at the source. A crucial factor which is affecting resource reservation is Quality-of-Service (QOS). The common interpretation of QOS is from a network perspective rather than a user or customer perspective. A more suitable view makes use of the two perspectives and yields two QOS characterizations (we can call them delivery quality and system QOS). A present  challenge is to identify the mapping from delivery quality to system QOS for a range of  system design parameters (e.g., data compression and network switching modes). User Traffic Characterization Although customers access the VOD system randomly, having a priori knowledge about  user access patterns can lead to a more efficient design. The system can make use of this information to manage network and storage bandwidths. As an example, if the traffic characteristics indicate that a movie is popular at a particular site, the system can replicate the movie locally to increase availability. The access pattern of users to the system will not be uniform over a given  24 hour period. Typically, one would expect the load to be low to moderate during the  daytime and to increase gradually through the evening and decrease again during the night. A hypothetical graph characterizing the access to a VOD database for a 24 hour period  is shown in Fig. 4. The access to the database is high during the evening hours, peaks at  around 9:00 PM, and is low-to-moderate during the day. This access pattern can be used for  designing schemes for various considerations like resource management; to update popularity tables, redistribute data, and reconfigure the system during off-peak hours. 0 5 10 15 20 time-of-day database-load Figure 4: A Schematic Daily-Access Model for a VOD System Similar models can be implemented and maintained for different geographical regions, movie categories, and individual titles. Such models are able to accommodate the differences in programming choices (e.g., children’s movies are more popular during the early evening hours) of different user groups. However, the complexity of these models, and their tractability is still to be established. Load Balancing An issue related directly to resource reservation is load balancing. The load balancing of VOD can be viewed as a combination of two sub-problems (i) The movie-storage  allocation problem and (ii) the resource location and connection establishment mechanism. Even though these problems are solved more easily individually, they are not independent  with respect to performance. From the perspective of a generic interactive  system, solving these issues is an open problem; however, simplifications can yield tractable solutions. As an example, if one assumes that a VOD system supports only stored data; i.e., movies  have to be digitized and stored before they can become available online, then the data  characteristics of a movie are well known in advance (e.g., the system has a priori knowledge  about the average bandwidth, burst rates, burst durations, etc.). This knowledge once available, can be used to simplify the design process. Making use of the metadata mechanism as described in Section 3 simplifies the task of management by decoupling the storage problem from the location problem.