Introduction to Multimedia Networks

	Before being overwhelmed by all the alphabet soup and jargon associated with teleconferencing and multimedia communications in general, network managers should realize that when all the networking issues are removed, the conferencing equipment breaks down into four types of components. Click here for the Components of Multimedia Conferencing.

Price declines and performance improvements have been enabled largely by advances in electronics and improved algorithms running on faster processors. The "plumbing" issues include the capability and availability of the pipeline, and standards. Here again, recent developments have altered the landscape. For example, the capabilities of modems for the analog telephone system have improved over recent years from 14.4 to 33.6 kbps for full duplex operations. Basic rate ISDN (128 kbps) is more widely available at ever lower costs, and T1/E1 lines are no longer rare oddities. Local area networks have improved too, with raw Ethernet performance moving from 10 to 100 Mbps, while switches are replacing hubs in many installations, providing further performance gains.

The definition and adoption of standards has accelerated as the platforms and network types used for conferencing have multiplied. A few years ago, all group and desktop videoconferencing products ran over ISDN, using the H.320 standard for audiovisual telephony over switched digital circuits. This standards compliance allows products to communicate with each other, and has made ISDN a proven videoconferencing pipeline all over the world.

For the public switched telephone network (PSTN), all videoconferencing products developed before 1996 used proprietary coding approaches and could communicate only with themselves. The same was true for LAN-based products. But in the past two years, the ITU (International Telecommunication Union, a United Nations body that sets international standards for communications) has adopted H.324 (for analog phone lines) and H.323 (for IP packet networks), and virtually all vendors now support these standards.

The rapid progress of packet switched technologies is now measured in "Internet-years." While a few years ago, nearly all of the conferencing solutions available to customers were based on circuit-switching technology, today many mainstream equipment suppliers have introduced network-based products that make it easier for LAN-attached users to conduct meetings at a distance.

In 1998, while price, performance, and plumbing issues for the network manager have left center stage, two others are moving into the forefront:

• The value proposition for multimedia-enabled networks
• The issues of interoperability and connectivity.

Videoconferencing speeds communications and helps employees work together and be more efficient. The real value is not in expense reduction, though expense reduction is usually the most tangible and quantifiable benefit, but in getting more things accomplished in less time with fewer people. Rapid communication is the real lifeblood of many businesses, whether it enables a parts designer to solve problems with a manufacturing engineer on the factory floor, a sales agent to interact with a customer across town, or a marketing manager to collaborate with a remote expert on the other side of the globe. The video-enabled network is thus the foundation for a better way to accomplish traditional tasks. Because these conferences are easier to do than in-person meetings, more of them get done, with more people, at less cost.

Another useful videoconferencing technology-based application is delivery of broadcast (one-way) audio-video streams. With a multimedia-enabled network, corporations can deliver to suitably equipped desktops and conference rooms live newscasts such as CNN as well as live internal broadcasts such as annual shareholders meetings, special messages from the president, or even new product announcements from the marketing department. The costs and network challenges are significant, but in cases where timely delivery of news is crucial, such as for Wall Street trading stations and financial services companies, networks supporting multimedia broadcasts are rapidly being installed.

IP-based telephony, another new technology, is also known as voice-over-IP (VoIP). For consumers, the allure is the possibility of using the Internet to make long distance calls for the price of a local connection. For corporations, the tangible benefit is also cost reduction by using the installed data network or intranet to carry voice traffic which would otherwise be traveling over the toll-based public telephone system. Many corporations have an extensive IP network already installed, and the ability to deliver voice (and other multimedia) services will evolve and improve as switching, routing, and MPOA (multi-protocol over ATM) breakthroughs are incorporated.

There are many technical challenges and tradeoffs facing the network manager with VoIP, but it remains one of today's fastest growing markets, perhaps because the cost savings are so direct and measurable. By some accounts, the average enterprise pays about eight cents per minute for internal telephone calling; with VoIP this can be reduced to less than two cents. For "greenfield" installations, there is even serious talk of using VoIP technology to eliminate the PBX entirely and have a networked server perform all the call control functions now allocated to dedicated switches. A related application with far fewer technical challenges is fax-over-IP. (Fax traffic in corporations is much lighter than voice traffic, but fax messages work far better in a store-and-forward-based system.) Lately, the ITU adopted a new standard for real-time fax-over-IP in which remote fax machines are connected over a packet-switched network and endure the latency and out-of-sequence packets problems.

A third class of applications facing network managers can be collectively described as the delivery of store-and-forward video. Whether the content is a recent speech by the chairman, a training tape on the latest product, an explanation of the company's health benefit plan, or a company-required education program on sexual harassment, the network issues are the same"bringing TV-quality audio and video to the desktop with VCR-like user controls for pause, rewind, and so on. A new class of hardware/software products called video servers with a wide range of functionality and performance are now coming to the market, designed for a range of network capabilities. Video servers make it quite feasible to "plug-in" store-and-forward capabilities to any multimedia-ready network. Many of these products are also looking at IP delivery mechanisms and are driving the interest in IP multicasting.

Achieving industry-wide interoperability is the focus of the International Multimedia Teleconferencing Consortium (IMTC), a non-profit corporation comprised of more than 150 member companies worldwide. The mission of the IMTC is to promote, encourage, and facilitate the development of interoperable multimedia teleconferencing solutions based on open international standards. IMTC activities include conducting interoperability test sessions; educating the business and consumer communities on the status, value, and benefits of the underlying technologies; and encouraging IMTC members to make submissions to standards bodies which enhance the interoperability and usability of multimedia teleconferencing products.

After bandwidth, which is deterministic, the next most important technical concern is latency, or end-to-end delay. In fact, delay (or lack thereof) is the primary determinant of user satisfaction levels. According to Madge Networks' Ron Nash, "The three primary concerns for a network manager planning on adding multimedia conferencing to the LAN are latency, latency, and latency. Basic bandwidth is usually available on the LAN, but end-to-end delay can cause an unacceptable conferencing experience for the end user."

Total latency is the sum of CODEC delays at the two end points and server/router delays for every network node between endpoints. Hence total latency is variable according to instantaneous network traffic. At 50 ms, delay is barely perceptible; at 250-300 ms, latency is annoying; at 600 ms, interactive speech is barely possible. For fast hardware-based videoconferencing codecs, compression/decompression takes about 225 ms. Another 50 ms are lost moving data around on the PC buses, so total endpoint delays sum to about 275 ms. Less capable codec designs may add another 100 ms to that figure, and software codecs may use up to twice that amount.

Network latencies, which must be added to endpoint latencies, vary widely. For H.320 videoconferencing on a circuit-switched ISDN connection, network latency is under 10 ms and constant. Quality of service in this case is a non-issue. On a packet-switched LAN, latencies under 80 ms are very common, but occasionally the delay can stretch far beyond this figure. Remember, there is no guaranteed quality of service on IP networks. Therefore, videoconferencing over the LAN using H.323 can produce very acceptable, business quality audio and video at 384 kbps most of the time. However, latencies of 500-2000 ms are not uncommon when communication is Internet-based and 5-15 hops (routers) between endpoints likely.

For users of network-based conferencing products, latency is the primary problem today. The IETF (Internet Engineering Task Force) and the IEEE are addressing these quality of service concerns with several efforts, including:

• RSVP (Receiver-Initiated Resource Reservation Protocol), which indicates to transmitting nodes the nature (such as bandwidth, jitter, or maximum burstiness) of the packet streams it wishes to receive. RSVP has been endorsed by most router and ISP companies. In fact it has been implemented by most vendors, but not yet turned on by most users.
  
  
• RTP (Real-Time Transport Protocol) is an application level protocol which provides end-to-end delivery of real-time (streaming) audio and video. RTP provides functions such as payload type identification, sequence number verification, internal time stamp transfer, and transmission data monitoring. RTP complements the functions of a lower layer network and is an underlying transport for H.323.
  
  
• RTCP (Real-Time Transport Control Protocol) is used for the control of RTP. RTCP monitors the quality of service and conveys information about the session participants. It periodically distributes control packets containing that information to all session participants.

Network managers need to protect the network from overloads which may result from overzealous use of LAN-based conferencing tools. This protection is the function of a new class of product called gatekeepers. While gatekeepers manage LAN usage, gateways are the critical component for tying the LAN to the WAN (and to the Internet) and are the network manager's best friend for managing bandwidth across this function. Depending on the number of LAN conferencing users, the WAN interface should be able to handle multiple ISDN-PRI lines, and to grow easily as users' needs evolve.

Conferencing will stretch resources in multiple directions. First, multimedia applications will need to interface with audio and video subsystems on end user stations. If your desktop computers are multimedia-enabled, you may have already crossed this bridge. Otherwise be prepared for dealing with Windows plug-and-play dynamics and IRQ conflicts associated with audio/video inputs and outputs. Second, network traffic will increase. Third, the demands of the user community will increase, since voice and video quality are very evident and very dependent on LAN/WAN performance. With traditional applications such as e-mail and file servers, performance is more tolerant of LAN delays, dropped packets, and so on. Finally, the need to communicate between endpoints on different LANs or to endpoints that are not LAN-attached will require smooth LAN/WAN intersections, marrying the traditional functions and responsibilities of the network manager with those of the telecommunications manager.

Network managers can optimize their chances of success by setting realistic expectations about multimedia conferencing. While downplaying the media hype, they should move aggressively in designated areas to set the stage for the next generation of network applications. The first step is to plan ahead, estimate the size and type of end points to be deployed, the rollout rate, and the ultimate end goal. This, like any networking project, can be critical to success. Simultaneous with good planning, managers should make sure that the existing network is in good shape and under control. This means understanding what the requirements are, and where the traffic loads occur. Typically, this means migrating to a switched network architecture as a first step. Then, a good place to get started is to attach a few conference rooms or group videoconferencing systems to the LAN. This offers the opportunity to test the waters with higher bit rate, higher quality conferencing systems, yet not burden the LAN with a large number of endpoints.

A first phase implementation should include at least one or two remote connections. While frame relay and extended IP networks are gaining attention, most customers today prefer ISDN for the WAN because switched data WANs are still more familiar and easier to design than packet-based switches.

With a WAN connection, you get great leverage from your initial multimedia LAN investment. Any of your multiple group systems on the LAN can connect to any ITU standards-compliant system practically anywhere in the world. The gateway handles all the technical issues. Your users can confer with colleagues, customers, vendors, or joint development partners. The savings come from reduced travel, but additional benefits accrue from easier communications and better relationship building. As users become more accustomed to the benefits of multimedia conferencing, and as bandwidth needs and usage patterns are documented, you can roll out desktop conferencing systems in a carefully structured program.