The digital age is now upon us. With millions of digital set-tops now deployed in cable networks, and thousands more being installed every week, network operators are beginning to look beyond simple broadcast services toward new, interactive services,

One service that many operators are banking on is video-on-demand. Because it offers the end user a method to control his television viewing habits, VOD is being pursued aggressively by all the major cable operators because this "personalization" of TV allows the viewer to watch his favorite programs anytime he wants to. Interactive, real-time control of movies, audio, time shifted-programming, e-commerce, games, banking and more will now be available to every digital subscriber instantaneously. This new model of TV consumption will reshape the way people live, work and play.

Figure 1: Video-on-demand system.

But before such services can be deployed, a network designer needs to understand VOD technology and make some fundamental decisions. Before evaluating a server solution, it's important to understand the components that are required to make a complete video-on-demand system, the technological issues that will be encountered during deployment, and server options that need to be considered before making an investment.

System components

A VOD system is comprised of four major components: the digital network/system architecture; a digital set-top box; a VOD server; and back office control software and hardware (see Figure 1). Before anyone jumps on the VOD bandwagon and starts testing or integrating a VOD server technology into the network, there are a few very important issues to consider. The first one relates to server options. Every server vendor will offer a wide variety of server and software options and features. Some server platforms are dynamic and can be configured with or without particular features, while others may be more rigid and less flexible as it relates to the architecture. Nonetheless, consider the system's performance, size, price, scalability and fault-tolerant features before making a decision.

Regarding system size, remember that as a result of the addition of Internet-based services, digital set-top box technology and telephony services, headend real estate and line powering have become scarce commodities. Therefore, it is highly recommended that the server manufacturer supply detailed system configurations and drawings showing rack layouts and powering requirements before implementing a server technology for any initial system launch.

As for scalability, make sure that the technology chosen is capable of being "easily and economically" upgraded as demand grows and digital set-top penetration increases. Also, make sure that system upgrades are modular so that they can be performed on site.

The rule-of-thumb today for deployment of an initial VOD system is to design it to meet a seven percent to 12 percent simultaneous usage rate, based on the number of digital set-tops deployed. An average number used today is 10 percent. That means that if 10,000 digital set-tops are forecast to be deployed, a VOD system should be designed to output at least 1,000 different streams simultaneously.

But it's also wise to plan for future growth. It doesn't make sense to integrate and deploy a system today that can't be scaled as demand increases. It may be wise to develop a five-year plan, based on digital set-top growth, that identifies all of the costs and additional equipment requirements for upgrading the VOD system each year.

Figure 2: Single broadband network.

Any good VOD system vendor will say its system is fault-tolerant, but fault tolerance has many definitions. Review them all carefully with the server provider. As the demand for interactive content delivery increases, so will revenues. Therefore, it should be of the utmost importance that a VOD system is completely fault tolerant. Evaluate the VOD system live in a working lab environment, and start disconnecting disk drives, modules, cables, power and anything else that the system may be subjected to in a real-world environment. Test it under every possible scenario once it's on line and running. Ask the server manufacturer for a copy of its automated test plan (ATP) so that it could be used for system verification and testing. There are technical issues and decisions to face before any commitment to deploying video-on-demand is made. Broadcasting digital signals (video, audio and data) into a broadband network requires the use of QAM (Quadrature Amplitude Modulation) modulators. To date, there are two levels of QAM modulation: 64-QAM and 256-QAM. The amount of (digital) bandwidth that can be sent through a 64-QAM modulator is 27.8 megabits per second (Mbps), while a 256-QAM modulator input bandwidth is 38.8 Mbps.

Furthermore, the level of encoding will determine the number of streams that can be sent through a modulator. For example, if an encoding rate of 3.85 Mbps is chosen, the maximum number of streams that can be inserted into a 256-QAM modulator will be 10 (10 x 3.85 Mbps = 38.5 Mbps). Of course, this equates to 10 streams per 6-MHz channel because each QAM modulator is equivalent to one channel on a broadband network.

Because of its larger input bandwidth capability, 256-QAM is becoming the standard format for digital modulation. Therefore, examples in this article use 256-QAM and an encoding rate of 3.85 Mbps for calculating and identifying bandwidth requirements. The encoding rate can be reduced in order to increase the number of streams that can be input to the QAM modulator, which saves bandwidth; however, as the encoding rate is lowered, picture resolution decreases. It is advisable to run tests to determine which encoding rate is best, keeping in mind picture quality, bandwidth and costs. Regarding system design, it should be decided early whether to design a system in a centralized or distributive format. There are pros and cons to both methods. This article assumes that the content being loaded onto the system is not pre-encrypted, and that the encryption (conditional access) is applied at the input of the QAM modulator just prior to being transported across the broadband network to the digital set-top box.

The centralized model (see Figure 2) requires all of the servers and back office hardware and software to be located in the main or regional headend. Loading and managing the content makes this approach very desirable. The negatives to integrating a centralized approach are cost, size, delay and most importantly, network bandwidth. A centralized design will require more floor space in the headend because all of the streaming servers, computers and QAM modulators are located in one centralized facility, which will also require a heavier power load.

Using the earlier examples of 10,000 digital boxes and a stream load of 10 percent (or 1,000 streams) at launch, the network's bandwidth allocation plan will have to be adjusted to accommodate for the 100 additional channels needed to deliver the VOD streams. Also, as digital set-top deployment increases, so will the number of streams needed, which will require additional channel bandwidth. Therefore, a centralized architecture has its limitations. Granted, there are alternatives to increasing bandwidth, such as dense wavelength division multiplexing or installation of QAM modulators into the hub sites. This latter method also requires additional equipment to convert the digital outputs of the server to an Internet Protocol (IP) or asynchronous transfer mode (ATM) format for transmission to the hub sites, where the IP or ATM is then reconverted to an acceptable input format for the QAM. This is a costly alternative and should be calculated into the overall cost of deploying VOD.

Figure 3: Distributed tiered broadband network.

Another consideration when deploying a centralized approach is delay. Depending on the overall size of the network, a subscriber located in the far reaches of a system may encounter delays that range from mere seconds to minutes while content is loaded into the set-top box. An alternative approach is the distributive model (see Figure 3). This requires that the main server(s), back office control software and hardware be located in the main or regional headend, while smaller streaming servers are installed at each hub site. This way, all content is loaded onto the server(s) in the regional headend. The smaller hub site servers store the top 20 or 30 "high demand" titles, as well as the first page (15 minutes) of the less-requested titles that reside on the main server(s). This way, a subscriber residing in the outlying reaches of the system who orders a low-demand title doesn't encounter any delay, because the first page of the movie will immediately be streamed to the set-top. The balance of the movie is streamed to the hub server, where it is stored and then seamlessly merged and sent to the subscriber without any visible interruption in service.

As for bandwidth efficiency, because the outputs of each server are fed directly into QAM modulators, the smaller hub servers will require fewer QAM modulators, which means less channel bandwidth allocated for VOD. Assuming that each hub server is required to stream 100 streams, only 10 QAM channels will be needed in each hub site. Under this configuration, the streaming load is spread out across the network via smaller content servers. An additional benefit to this method of deployment is cost, in that powering and space requirements are spread throughout the system and additional equipment to convert and transport digital outputs from the VOD server(s) isn't required.

The downside to this method is that content needs to be propagated and loaded onto each hub server. Deciding which format to use will depend on the design and layout of each individual system, as well as the features and options offered by the server vendor.


Before making a decision on which VOD server technology to deploy, request a system integration plan from the server vendor. Evaluate all server platforms available today. As a final option, plan to run test trials with the competing server technologies to determine which technology and system architecture best meets the needs of the subscribers.