The Survivor's Guide to 2004: Converged Voice, Video and Data

As a result, many companies are hedging their digital-convergence bets. Movement toward a converged IP network for voice, video, and data will remain limited though 2006, according to Meta Group. Two reasons cited are a lack of demonstrable ROI and management tools. Sure, whenever you back a new technology, there's a danger that the benefits won't outweigh the burdens. All bets involve risks, but if you recognize and manage those risks you can reduce the potential for loss.

The key factor to consider is what's at stake if you don't converge: You'll continue to manage and support disparate technologies to generate and communicate enterprise content. And you'll not only have to manage and maintain separate voice, video and data networks, you'll also have to create content in multiple formats to deliver to employees, clients and partners.

Speaking of costs, convergence can save you money on network connections. Most enterprises have an IP network infrastructure with leased lines--T1 or T3--to connect branch offices and distribute multimedia content. A T1 may cost $900 per month or more, depending on location, according to the TIA 2003 Telecommunications Market Review and Forecast. Most enterprises also support a PBX or lease a Centrex system for analog access. Then there's the price of videoconferencing equipment and a monthly lease for video transmission over ISDN. Setting up an ISDN line for videoconferencing carries a high installation cost--$500 to $800. And if it's a long-distance line, monthly service charges can range from $60 to $150 per month. Add in expenditures for postage and couriers, and you'll see it costs a bundle to communicate and distribute information. Enterprises that achieve such communication cheaply and efficiently while maintaining high-quality content are going to come out ahead in sales and revenue.Think next year might be the time to embark on a convergence plan? Here are the key business elements and a rundown of potential risks.Identifying the big-picture business case for convergence is easy: Reducing the number of networks you maintain and manage will cut operational costs. For example, you won't need analog line cards in your PBX, and you'll be able to eliminate separate voice cables to desktops. Moving and/or adding employees on one voice/data network is cheaper and easier than doing so across separate networks. Sending voice calls over the data network reduces your reliance on public carriers and your monthly phone bills. And you can decrease the cost of owning and managing separate networks and let your IT talent focus on the benefits, rather than the burdens, of networking.

Convergence also provides rich applications that merge voice, video and data content, like streaming media and videoconferencing. These converged apps can add innovative and revenue-generating enhancements, such as video chat rooms and "click-to-talk," to your CRM and ERP systems. They also can lay the groundwork for instant messaging, enable unified messaging and leverage high-quality digital materials.

Convergence can even reduce your overall bandwidth requirements via innovations in voice and video compression. When done right, VoIP systems can compress voice data down to a fraction of the original bandwidth required for an analog call and still maintain quality.

Business Risks of Convergence

But data networks are not like voice networks. Most enterprise networks use Ethernet and TCP/IP--architectures in which packets can be late, become distorted or even get lost. These network problems don't unduly hamper data applications: Ethernet hardware and IP mitigate many of these concerns on busy networks. But for time-sensitive voice and video packets, quality is reduced if not delivered in real time with minimal packet loss.Quality indicators for converged networks will be familiar to most network administrators. They are the usual suspects for bandwidth problems: network delay, jitter and packet loss. For 2004, to manage risk you must overprovision bandwidth or implement QoS (Quality of Service) techniques (see "The Unsavory Characters in IP"). Remember, though, that data networks simply are less reliable than the PSTN.

Then there's the personnel problem: There's not a whole lot of expertise out there in managing and administering converged networks. To date, most enterprises have relied on separate departments to manage voice, video and data. Bringing these camps together raises management and support issues that may require retraining for your IT staff.

To start the year off right, maintain the good common sense you used to put your data network together: Proactively manage the network and monitor your bandwidth requirements, and stay with best-of-breed solutions that adhere to open standards and architectures. This will let you develop a plan to leverage your network infrastructure and integrate new converged platforms, such as IP-PBXs and voice gateways. And don't get locked into a proprietary system, which will limit your future choices.The TCP and IP protocols were designed for terminal access to remote systems, file transfer and e-mail. They aimed to establish cost-effective connectivity over a wide variety of devices and provided a "best effort" to deliver packets. TCP/IP's best effort, however, is not sufficient for voice traffic that demands a high QoS. Key for 2004: Make sure you have a QoS strategy. But don't work in the dark. Your first step should be to take account of your current network utilization with an eye toward convergence.

Cisco Systems and other infrastructure vendors have tools to measure bandwidth throughput for devices. VoIP and network-management vendors also address this issue. For example, NetIQ's VoIP Assessor simulates a VoIP implementation to determine if your network has what it takes to support the load. ITWorx's free NetCelera Scout classifies and reports on network traffic. With these, you can scope out the network and remove the problems that may lead to network delay, jitter and packet loss.

You also need to calculate in the application delay for VoIP caused by the audio compression and decompression. Although audio codecs like G.729 can reduce the audio data rate as low as 8 Kbps, the compression to send it over the network and decompression to listen to it can add to delay--the G.723.1 codec is especially bad in that regard. Make sure your codec does an efficient job of compressing and decompressing audio. For efficiency and to reduce overall traffic with voice over networks, you can combine multiple audio frames into one packet and use silence suppression. Or manage bandwidth by prioritizing traffic--that is, give audio traffic priority over other streams.Apply 802.1p priority at Layer 2 of the OSI model or DiffServ (RFC 2474) service tags at Layer 3. Then, once packets are labeled, act on them as they pass through supported devices using a prioritized queuing mechanism like WFQ (Weighted Fair Queuing) or CBQ (Class-based Queuing), or establish a specialized route using RSVP (Resource Reservation Protocol). Priority settings can be applied in the end user's system if the application supports it. If not, they can be applied in multilayer switches and other devices, such as packet shapers from the likes of Packeteer or Allot Communications. But remember, prioritization schemes do not guarantee delivery. There won't be a new tool in 2004 that will guarantee TCP/IP traffic over the network. But the devices out there to optimize network performance are more mature and widely supported in network infrastructure.

We don't recommend swapping a perfectly good legacy telephone system for VoIP. Most companies considering VoIP are moving into new quarters or have an aging phone system that must be replaced. If this describes you, look closely at VoIP, or at an IP Centrex system from a service provider (see "Centrex, IP Style"). VoIP systems tend to be easier to administer than IP Centrex because they're more user-friendly.

To ensure your implementation is up for the long run, carefully consider SIP (see "It's Time To Take a Look at SIP"). SIP has emerged as the successor to H.323 to deliver a standards-based VoIP phone system for enterprises. This will impact other real-time communications technologies like instant messaging, presence management and unified messaging and facilitate delivery of voice over 3G wireless and cable networks. And like H.323, SIP will make it easy to add voicemail, IVR (interactive voice response) and "click-to-talk" features. With a simple infrastructure consisting of SIP proxy and registrar servers, you can orchestrate call requests to SIP-compliant devices and manage presence.

When buying SIP devices, watch what vendors do rather than just listening to what they say. Many VoIP suppliers, including Alcatel, Cisco, Interactive Intelligence, Mitel and Siemens, are touting new SIP-compliant products. Even Microsoft has incorporated SIP into its XP Messenger client on every XP desktop. Make sure the devices you pick adhere to the SIP standard.Because the demand for videoconferencing products has been sluggish, according to Frost & Sullivan, competitive pressure on vendors could lead to some good deals. Streaming video and videoconferencing are effective tools to enhance communication and collaboration, provide in-house training and increase sales. One key factor in selecting a streaming media system is the encoding scheme or bit rate. Bit rates for encoding video determine your bandwidth requirements as well as the quality of transmission (see "Encoding Schemes" page 62).

Which video server solution you choose depends on the quality of the video and your audience's bandwidth requirements. If you deliver video on your enterprise network, an MPEG-2 or -4 server will do the job. If you plan on sending content over the Internet, you should select an efficient codec, such as MPEG-4, and look at products that support a strategy for DRM (digital rights management) as well as support for wireless handheld devices like 3GPP (Third Generation Partnership Project). A new standard to create, deliver and play back multimedia over high-speed wireless networks, 3GPP uses MPEG-4 to optimize video delivery to wireless handheld devices. And don't be afraid to look at proprietary solutions from Microsoft or RealNetworks that deliver quality video ranging from 300 Kbps to 1 Mbps. Decoders or players for end users are usually free.To take the first steps toward videoconferencing, you don't need a system sporting full-motion video at 30 frames per second. Meetings that feature "talking heads" require only 10 fps to 15 fps, and these rates can be accomplished using a desktop video camera with software supporting the H.323 IP video standard, such as Microsoft's NetMeeting, or a SIP-compliant endpoint, like XP's Messenger. For as little as 128 Kbps per session, you can set up one-to-one meetings between desktops to test the waters--not much of a gamble here unless you have a lot of bad hair days.

For both desktop and room systems, you will need an MCU (multipoint conferencing unit) to bring three or more conferencing systems together into the same conference. An MCU manages the audio, video and data to and from each conference participant. Audio from all endpoints is mixed and delivered to each endpoint in full-duplex. Before you buy an MCU, gather your requirements for performance, and scrutinize the cost and the upgrade path to advanced feature sets. Check how users implement any existing conferencing system. For example, do they collaborate on projects and have a need for application sharing? MCU vendors, like VoIP providers, should be steering toward an open systems architecture and have a road map for SIP support.

Sean Doherty is a technology editor and lawyer based at our Syracuse University Real-World Labs®. A former project manager and IT engineer at Syracuse University, he helped develop centrally supported applications and storage systems. Write to him at [email protected].

Post a comment or question on this story.

MPEG (Motion Picture Engineering Group) needs no introduction. It has been an industry standard for audio and video compression since MPEG-1 debuted in 1991. With a bit rate of 1 to 1.5 Mbps, however, MPEG-1 is not known for its video quality. Its successor, MPEG-2, is more commonly used in the enterprise for high-quality training videos and corporate communication. MPEG-2 needs a fat pipe to satisfy up to 5-Mbps bandwidth requirements. On the plus side, it provides variable resolutions and bandwidth that enable playback on computers, standard TV and even HDTV (high-definition television). If you are using an MPEG-2 video server, 2004 is the time to look at MPEG's newest codec: MPEG-4.

Like MPEG-2, MPEG-4 has variable resolutions and bandwidth requirements for multiple uses. The MPEG-4 bit rate can go as low as 5 to 64 Kbps for low-resolution video for mobile users. At low resolutions and bit rates,it can replace MPEG-2 and provide better-quality video using the same number of bits. For high resolutions, it can ramp up to 2 Mbps for film-quality video over enterprise networks. If you think your MPEG-2 server is providing sufficient quality, an MPEG-4 server--like VBrick Systems' VBXcast, Starbak's Torrent OSA or even Apple's QuickTime Streaming Server 5 (built into OS X)--can give you the same quality but use less bandwidth.

And if bandwidth is a concern, you should look at other software that can deliver high quality over impoverished networks, like Microsoft's Media 9 Series or RealNetworks' Helix Universal Server. MPEG-4 also can better manage and protect enterprise content owners.Akamai Technologies: EdgeComputing improves access to Internet applications and data for enterprises.

Alcatel: Makes the cut for its commitment to SIP and ability to offer a hybrid system.

Apple: QuickTime Server bundled with with OS X includes MPEG4 and 3GPP support. Get the open-source version (Darwin) for free.

Interactive Intelligences: Dark-horse VoIP supplier with SIP-compliant servers and UM systems.Microsoft Corp.: Windows Media 9 Series bundled with server license with DRM support; SIP-compliant XP Messenger client and RTC (Real-Time Communications) Server worth checking out.

Polycom: Cost-effective VSX 7000 set-top videoconferencing system and customizable, SIP-compliant SoundPoint IP 600 phone both caught our eye.

Radvision Corp.: ViaIP videoconferencing systems play nice with Microsoft XP Messenger and Office.

3Com Corp.: EDS will include 3Com switches, routers and VoIP products in its offerings.

VBrick Systems: EtherneTV Media Distribution Server delivers live and stored video to TV and PCs. "Fear Factor," anyone?VCON Visual Communications: Media Xchange Manager version 4 is a video PBX with gateway support for SIP and H.323 endpoints. Did we mention SIP?

Zultys Technologies: Notable for pure SIP-based PBX appliances and phones.

• Network Computing's digital convergence white papers and research reports

• "Does QoS Deliver?"

• "SIP Packs a Punch">"SIP Packs a Punch"• "Delivering Content to Handhelds"

• "Buyer's Guide: Choosing a VoIP PBX"

• "Come and Get It"

• "VoIP Wake-Up Call"

If network latency is high, voice conversations can be painful at best. TheITU-TG.114 standard for one-way transmission time recommends approximately 150ms for maximum one-way latency to maintain voice quality. In someimplementations, as much as 400-ms delay can be acceptable. That is, if the badguys of VoIP (voice over IP)--echo, jitter and packet loss--don't rear theirugly heads.

Echo: Delays in the round trip between caller and receiver can lead toechoes on the line. Echoes occur when voice signals are reflected back to thespeaker and are caused by impedance and data conversion between the telephoneset and the network. If the round-trip delay exceeds 30 to 50 ms, echo mayreduce the system's usability. Echo cancelers can ameliorate the problem, butthey're not foolproof. A canceler has a limited amount of memory to compare areceived voice pattern with the current voice pattern. When they match, thecanceler cancels the duplicate as an echo. But where the network delay betweenthe caller and the receiver is excessive, a canceler may not have sufficientmemory to capture, compare and cancel a potential echo.Jitter: Jitter is the variance in delay caused by different endpoints onthe network. When networks suffer congestion, device buffers fill, andtransmission can be delayed. This leads to a variation in the time a packetshould arrive at an end node and when it actually does arrive. And jitter causesproblems when decoders play back audio and video. One way to deal with this isto adjust the jitter buffer on devices. As audio packets are received, they passthrough a buffer, where they're stored before playback. If incoming packets aredelayed, information in the buffer can be played until empty or the delayedpacket is received. Although the buffer can be increased as jitter gets worse,it introduces a "hard" delay in playback and can degrade the quality of thecall. And if packets are delayed longer than the time set for playback, they'rediscarded. It makes no sense to play back a packet out of order for a voiceconversation or a video file.

Packet loss: In cases of high network congestion, packets may not only bedelayed, they may be lost or dropped. If your network is losing more than 10percent of packets due to congestion, voice quality will suffer. But note thatif packet loss occurs randomly over a length of time, it may not be detrimentalto call quality. The ear is not able to detect small packet loss duringvariable-pitch speech. If the packet loss occurs in traffic bursts, however,you're likely to lose a number of packets simultaneously, and that will resultin long delays in speech, where words and even sentences could be dropped.