Spread the news: Video-on-demand is real. Since the beginning of this year, any digital cable TV subscriber of Time Warner's Oceanic Cable division on the island of Oahu has had access to VOD. Fifty thousand Scientific-Atlanta Explorer set-tops have been deployed over a base of 256,000 subscribers, and hundreds more are being installed every week. Customers can order any movie from the selection menu, retaining complete video stream control that allows them full VCR-like functionality such as pause, fast forward and rewind. Many other interactive services such as banking, karaoke, local news replay, educational services, Internet access and e-mail are in the test phase. Pizza, however, can be ordered currently from selected areas simply by using a remote control and following an interactive user interface on the TV screen.Planning stages
Planning and provisioning for VOD and other interactive services have been an essential part of any Time Warner cable system upgrade since the early 1990s. Corporate specifications such as node size (500 households), frequency bandwidth (5–40 MHz return/50–750 MHz forward), hub size, fiber count, optical and RF performance specs, etc., were established a decade ago with great foresight. They were validated by the relative ease of implementing interactive services such as analog and digital IPPV, cable modem, and most recently, VOD services.
In 1998, advances in server technology, software development and deployment of digital set-tops prompted Oceanic Cable to take a serious look at the deployment of residential VOD services. On one hand, the pricing of system components was approaching economical levels, and on the other hand, digital MMDS services were exerting competitive pressure.
The vendor evaluation and selection process was completed in April 1999. The MediaHawk server (from Concurrent Computer) with Prasara Technologies back office software, was selected as a cornerstone of the VOD system, overlaying existing Scientific-Atlanta digital headend hardware and software. Installation of servers, QAM modulators and back office software began in July. Service from the headend to the employees islandwide was available within a month. Streaming capacity was increased by installing servers in the hubs, allowing staged release by regions in December. The service was opened to all subscribers in early January 2000, followed by a successful demonstration of VOD functionality to the press and major program providers in February.
The VOD trials and deployment at Oceanic Cable were based from the beginning on solid economic ground:
- There were no slush funds available for the exploration of technologies that would produce, at best, academic results.
- The plan called for utilizing existing infrastructure (e.g. digital headend, hubs, video and data networks, billing system, etc.) to its maximum potential.
- Existing human resources had to be utilized. To date, there has been no hiring associated with the VOD deployment and there is no person with duties dedicated exclusively to the VOD services.
- Time was of the essence. The goal was to succeed while not dwelling on minor discrepancies. For example, a less than perfect-looking user interface does not affect service reliability and could be improved in the future.
- Technology was available. Its performance and economic viability had to be proven in a large-scale deployment.
- Viewer interest existed. It was time to move beyond modeling and testing, and launch a competitive and profitable service
A VOD system is typically comprised of five major components: the video server; the headend; the back office control software; the network; and the digital set-top box.
The video server stores, manages and delivers the VOD content. The headend and hubs provide the RF managing, processing and modulation for the delivery to the broadband network, which in turn delivers the signal to the customer's set-top box. A control section of the network interconnects all of the devices. The complex software environment, consisting of digital headend operating system, server, client, middleware, administration, application, network management and back office software, is typically provided by non-related vendors. The requirement for flawless interoperability of all of these software components often represents the biggest hurdle to integrating and operating a VOD system.
The back office software performance will have, in the long term, the largest effect on the recurring costs for the VOD operator. The non-recurring costs consist primarily of the initial investment in the hardware systems and components. Equipment that is shared by other broadband applications include the set-tops, basic digital headend and hub infrastructure, the IP network and the HFC distribution network. The cost of servers, QAM modulators and content transport network remains as the main component of VOD capital expenditure.
The cost of servers is typically given as a cost-per-video stream, rather than a cost per storage capacity. Today, it is quoted typically between $300 to $400 per stream. One must be cautious, however, to understand the program stream bandwidth. A narrower bandwidth stream should cost less.
The cost of QAM modulators (at 256 QAM modulation setting and depending on the MPEG-2 encoding rate, each typically supports 10 streams in the 38.8 Mbps bandwidth) has come down recently. The modulator cost, however, still contributes significantly ($150 to $200 per program stream) to the overall cost.
Unless one chooses wisely, it is the cost of the network that will represent the lion's share of the VOD deployment budget. The selection of VOD architecture (centralized or distributed) and the selection of the associated transport network represent the largest single controllable capital expense item in the VOD deployment budget, and as such, deserve a closer look.
There is no optimum solution that would apply equally well for all systems that are considering deploying VOD. There are some common questions, however, that a system should include in the decision-making process. The answers will help define the optimum design and assist in the vendor selection process.
Traditionally, the consensus was that VOD is a centralized service. Most people believed that video servers would be located only in large, regional headends because of cost, size and maintenance issues. The content would then be distributed using real-time streaming via high-speed data networks (Sonet, ATM). As the cost and complexity of servers were reduced to a degree, locating a large capacity server at the system's main headend became a widely accepted model.
When Oceanic Cable began accelerated planning for residential VOD in 1998, the central server option was explored. The size of the system (340,000 homes passed) called for a capacity of 8,000 video streams. At 3.85 Mbps per stream, we required approximately 31,000 Mbps streaming capacity from the headend to 15 hubs dispersed throughout the island of Oahu. The MPEG-2 digital format does not map efficiently into an ATM cell, however, so an even larger ATM network was required. ATM to DVB-ASI protocol converters further added to the cost and complexity. Using such assumptions, large-scale deployment of video services was not an economically attractive proposal.
Further research and review of advanced technologies revealed intriguing alternatives, including dense wavelength division multiplexing (DWDM) technology; and DVB-ASI transport from a central server to the QAM modulators located at the hubs.
Both methods remove the need for protocol conversion that is required for ATM links. While cable industry applications of DWDM were in the developing stages, a suitable DVB-ASI transport product could not be found. The nearest match was a D-1 digital video card, a subsystem of ADC Communications' DV-6000 fiber transport product line. Oceanic was already utilizing the DV-6000 backbone transport in other applications, and had excess capacity available. Oceanic asked the manufacturer to develop a lower cost DVB-ASI interface card. The product now serves as a primary means of real-time video streaming from our library server at the headend to QAM modulators at all hubs.
Other factors that affected the final VOD architecture came as a result of several criteria, including:
1) Existing headend and hub space availability. Headend space was at a premium and not easily expandable, while the hubs had ample space. Installing a few hundred QAM modulators in a single headend, or creating a couple more super hubs, was not feasible.
As the DWDM transport alternative requires the modulators to be co-located with the centralized servers, the space constraints at our headend eliminated that option.
2) Existing network architecture and its utilization. Analog and digital video and audio services were delivered to and from the headend to all hubs via a fiber AM supertrunk. On a temporary basis, some of the available capacity is now used as a VOD backup link.
Cable modem and institutional data network transport services utilized an ATM OC-12 network. The same network provided IP connectivity between Scientific-Atlanta digital service components. We extended that connectivity to the VOD equipment for the control purposes, and 100Base-T video file propagation* functions from the headend to all hub servers.
The DV-6000 digital fiber backbone was providing IPPV data, telemetry, inter-office telephony and tributary video feeds. The high-speed network is currently divided into three regional rings that congregate at the headend, giving it an aggregate capacity of 7.2 Gbps, offering ample access capacity. The capacity can be further multiplied, several times if needed, by adding DWDM components. With the availability of the native DVB-ASI transport card, the ADC network became a backbone of the VOD transport mechanism.
3) VOD "readiness" of the headend and hubs. Assuming that good engineering practices have been reasonably followed, and digital services have been operating reliably, VOD components should integrate with an existing system without much problem. You do want to make sure, however, that your RF management system is ready for the task. With so many other technical uncertainties that an introduction of a complex new service brings, a prudent engineer would want to virtually eliminate the headend and hub RF network from the list of potential problem areas.
When cable modems were deployed a few years ago, we realized that building forward and return signal combiners and dividers from components available on the market will not do in the long term. Constant changes in modulator and demodulator groupings, addition of new services and new serving areas, reliability, isolation and space requirement problems, etc., have prompted us to look for a more permanent solution. The goal was to make complex combining and narrowcasting groupings available for new services, while not disrupting currently active services.
At that time, no suitable product was available in the market, until we found VisionTeq. Following our original conceptual design and improving on performance specs, VisionTeq delivered highly integrated, compact and reliable forward and return combiners/dividers. The grouping and narrowcasting capabilities of the equipment enable quick and non-disruptive configuration changes at an affordable cost. The expanded product line has been implemented in a number of systems.
4) Reliability issues. In a complex system such as VOD, there are many opportunities for failure. Our previous experiences with video servers (SeaChange used in our ad insertion and hotel VOD services and ASC servers used in our program time shifting applications), made us confident that modern servers with failsafe capabilities can run unattended for long periods of time. Locating servers at unmanned hubs was therefore feasible from the maintenance and reliability point-of-view.
A careful design of system architecture will minimize the possibility of service outages. In our system, any hub server or QAM modulator outage (failure, maintenance, upgrade, etc.) would not cause denial of new VOD orders because alternative resources are available from the headend. Extensive remote monitoring and system status reporting procedures are also available.
5) Content format and availability. While a modern cable system successfully manages more than 100 satellite and local programming feeds, movie content suppliers are greatly concerned about the security of MPEG-2 encoded files; therefore, establishing a controlled environment for the delivery and loading of these files (typically DLT tapes) is essential. In the future, content files will be automatically downloaded from a satellite feed to the servers. The system design must accommodate current relatively crude downloading methods and allow for the future automated process.
The encoding format is typically MPEG-2, main level, main profile with a specified video and audio encoding rate. There are many finer points, however, that must be included and conveyed to the encoding houses in order to provide the content format compatibility with the operator's system. Time Warner's corporate office provided these specifications to assure that once a program is encoded, it will play without difficulties at all divisions that will implement a VOD system.
6) Content asset propagation, ordering, billing and royalty payment management. This topic is so vast and important that it warrants a dedicated article. Performing such functions manually would prove to be labor-intensive and cost-prohibitive for any large system. Therefore, an automated VOD back office software management system must be compatible with the operator's legacy billing system. This fact in itself may be a decisive factor in selecting the VOD system vendor.
Unlike most other cable system operators that rely on outside billing service providers, we developed a sophisticated, in-house billing system. As Noreen Kuniyuki, director of information systems, and a driving force behind much of the project can attest to, having an in-depth understanding and management control over our own billing system greatly expedited the integration of the VOD back office software suite.
7) System documentation. Video content will be delivered to the customer only if the system knows which QAM modulator streaming resources are available to that customer's set-top box. Do you keep a database that defines which serial number set-top box is at a particular street address that is served by a specific node? What hub and laser transmitter is that node fed from? These are the parameters that define TSID (Transport Stream Identifier) which is instrumental in routing the desired program file from an available resource to the specific QAM modulator. The set-top box will need to know which channel to automatically tune and acquire only the requested program stream from that channel. Being able to manually create a headend and hub TSID table and correlate it to an accurate set-top box distribution database was one of the main contributing factors to our early VOD deployment success.
If your system is like most others and does not maintain such a database, do not despair. Help is on the way in the form of an auto-discovery mechanism. For example, software enhancements will shortly allow Scientific-Atlanta set-tops to find and define the QAM modulators it can be served from. This is a major milestone and will open the VOD deployment option to many more system operators.
When the 750 MHz HFC upgrade began six years ago, Oceanic Cable began a tedious process of isolating and mapping the plant into serving areas of about 500 homes passed. Each serving area is served by a fiber node, and as the system expands and new nodes are added, the database needs to be updated. Every time a new set-top box is added or deleted from an existing serving area, the database will dynamically update its status as well.
8) Personnel skill levels. VOD is just another complex service offering for which it pays to learn how to walk before we run. For example, having experience with successful cable modem services or cable telephony deployment is a definite advantage. The experience will reduce the number of issues that would otherwise be often wrongfully allocated under the VOD problem umbrella.
Having your own personnel trained in the data networking and digital headend maintenance procedures ahead of the deployment is a must. Having a person with the video server hardware and software maintenance skills is a plus and will reduce dependency on the equipment vendor. Understanding the back office suite and the digital cable system operating software interaction and integration is a must. The knowledge will assist in daily operations and occasional system problem resolution tasks.Network system architecture
After a thorough examination of all the alternatives, a hybrid system of network topology and content storage hierarchy was developed. It could be described as a distributed tiered native DVB-ASI network system architecture. The term "distributed" describes the fact that the servers are distributed to all hubs, with the exception of the smallest ones. Those are served directly from the headend or the nearest large hub. The term "tiered" denotes the fact that the content is divided between the high demand content stored in servers located closest to the customer (in the hubs, for example), and low demand content stored on a larger server that is centrally located.
For the sake of this example, we could load 100 of the most current and highly requested movies on each of the hub servers. At the same time, we could store 500 archive-quality movies, karaoke files and some local news content on the central or library server that also contains the 100 high demand movies. Let's also assume that 80 percent of the requests will be for the popular titles. Therefore, only 20 percent of the total orders would require resources of the broadband transport network, significantly reducing the total implementation cost. As the servers are priced on a per-stream basis, they will cost the same amount of money whether they sit centralized at the headend or distributed between hubs.
The term "native DVB-ASI" denotes that this is a high-speed digital video network that transports MPEG-2 multi-stream signals in its original form with no protocol conversion. Originally, we were not able to find a network that would be available commercially with such characteristics. Manufacturers built the required elements per our request and expanded into a new product line. The network requires no switching components. Its cost and reliability compare favorably with any ATM network. The payload efficiency is 99 percent, versus much lower efficiency for ATM networks. No protocol conversion is required; the DVB-ASI port from the central server is virtually transported to any hub. Streaming is straight to the QAM modulator in the hub; no caching at the hub server level is required.
The term "fault-tolerant" denotes the fact that each of the set-top boxes on our system has access to the resources from a number of QAM modulators and at least two servers. Our idea of a special resource allocation algorithm protocol that will balance the load between available resources and also bypass the faulty elements has been implemented on our system by Concurrent Computer Corp. For example, if a hub server or the hub QAM resources are not available (too many orders or a faulty condition), the resources will be available from the central server via the DVB-ASI network or, to a more limited extent, via our AM supertrunk.
The VOD system has an initial capacity of 3,200 streams. The system has a sophisticated order and traffic analysis reporting system that can be used for marketing and engineering purposes. For example, after analyzing peak take-rates from all the hub areas (or even nodes if required), we can allocate the new streaming resources to where they are most needed. Building a network to its full size (based on purely theoretical projections) from the beginning is costly. A more prudent approach is to expand the capacity using trend analysis based on real peak demand data.VOD revenues are for cable to lose
There is no doubt in the author's mind that the hybrid fiber/coax cable infrastructure currently represents the most cost-effective, two-way broadband pipe to homes. When properly engineered, its narrowcasting capabilities can be fully utilized for interactive services such as video-on-demand. Customers crave high-speed interactive services and are willing to pay for them. Digital video services are popular, yet for some operators, it took the threat of competition from DBS and DSL before the investments in upgrades and new technologies were made.
There is no time to waste in implementing VOD services as well. Competitive VDSL (Video DSL) services are appearing in some areas (e.g. Phoenix). With newer video compression technologies such as wavelet and MPEG-4, a more acceptable video quality streaming via Internet is not too far away. Combined with caching of digital video content that is currently available at the consumer level (Tivo, ReplayTV), satellite delivery and early program release windows (Blockbuster), the competition is more credible than ever.
VOD deployment at Oceanic Cable has proven that the technology is here and available at prices that support a solid business plan. VOD is no more difficult for an operator to implement than cable modem services are, and if engineered right, VOD is, in comparison, also more reliable.
The services are in demand, and the race for customer loyalty is for the cable industry to lose.
Questions? Contact Lebar at (808) 625-8408