2003 Survivor's Guide to Digital Convergence

This will be the year to come to grips with network delay, jitter and packet loss, implementing network QoS and applying content-delivery technologies.

December 6, 2002

15 Min Read
Network Computing logo

CTI (computer-telephony integration) equipment also came into more widespread use as enterprises invested in call-center technology to attract and retain customers in 2002; that category is projected to rise by 13 percent per year (CAGR) from 2001 to 2005. And as more employees work remotely over wireless and broadband technologies, such as cable and DSL, digital convergence will bring them all the benefits of the office while they're on the road or at home (see "U.S. Broadband Services").

Unified messaging solutions also continue to grow, with a projected 32 percent CAGR from 2001 to 2005--further evidence of this category's viability.

Not all enterprises have found the reasons for convergence compelling enough to warrant purchasing new equipment to replace older, but working, equipment in the near term. Enterprises with PBXs and videoconferencing systems in production may not find a sufficient incentive to migrate these services to their IP networks in 2003. Your systems may still have features you've never seen. And they may still be depreciating in value. Replacing them and writing them off would not sit well on your bottom line. But if you need to replace legacy equipment, if merger takes your enterprise beyond the maximum limits of support, or if you're relocating to a new building, you must invest in technology with IP support.

Whether your enterprise is on the road toward convergence or still in the planning stage, you need some assurances that converged voice and video applications over IP will benefit customers and employees and not burden them with long delays and intermittent network outages. When convergence happens, your IP network must be ready.

Survivor's Guide to 2003



Mobile & Wireless
Network & Systems


Storage & Server


Business Apps

Digital Convergence

The Business Case


Convergence: Ready or Not

IP-enabled equipment and applications have become a dominant force on the Internet and in Web services. IP provides an adequate transport over Layer 2 network protocols, such as ATM, Ethernet, frame relay and ISDN, as well as in Layer 1 networks using ADSL and cable. IP's supremacy is rooted in the simplicity of its best-effort attempt to deliver a packet to its destination. A best-effort protocol, however, may not be adequate to deliver real-time voice and multimedia packets.

Before you consider convergence, you need to evaluate your network's service levels. We recommend using four indicators: bandwidth, delay, jitter and packet loss. If your network is barely meeting its service levels, you'll need to provision more bandwidth and consider distributing the available bandwidth using subnets, VLANs (virtual LANs), traffic-prioritization equipment and content-delivery technologies to help you meet service levels and keep voice and data traffic flowing. In general, bandwidth should be as large as possible while minimizing delay, jitter and packet loss.

Bandwidth Needed for Common Coding Algorithimsclick to enlarge

When you apply this rule in the real world, remember audio and video traffic is primarily transmitted over UDP, which, unlike TCP, does not retransmit lost packets. In the new network, the show must go on, and there's no place for any dropped or out-of-sequence packets. Long network delays result in stopgap conversations with extensive silent periods between sentences. Network jitter, the variation in end-to-end delay and sequential packet delivery, results in jerky video and stuttering audio. The same result occurs when intermediate devices drop or lose packets because of congestion.

The ITU (International Telecommunication Union) standard G.114 recommends that network transmission time should not exceed 150 ms (milliseconds), including the delay in equipment processing time and the propagation delay in traversing the network. In the real world, satisfactory performance for audio and video continues with up to a 300-ms delay, but as 400 ms approaches, quality deteriorates. If network delay is a problem, identify the congestion points and segment them into subnets or VLANs to share the bandwidth. Or provision more bandwidth and add the capacity necessary for convergence. So how much bandwidth is enough? Envision an enterprise with switched, 100-Mbps connections to the desktop. This is ample room for an employee to pick up an IP phone, attend a videoconference, view an MPEG-1 training video and step through the business processes of a CRM package simultaneously. But such data-intensive applications would be unusable if employees needed to access them over the 1.54-Mbps T1 links still found in most organizations. Let's consider just the requirements for voice, videoconferencing, and streaming media from this example.

Voice-traffic bandwidth depends on the coding algorithms, or codecs, used to convert analog voice waveforms to a digital stream, and typically range from 24 to 80 Kbps. Although these amounts may seem minuscule, they add up when you consider the number of simultaneous calls on a network.Videoconferencing systems, which traditionally used dedicated ISDN lines, require six data channels (or B channels) at 64 Kbps. That translates to 384 Kbps, though this requirement can vary. Talking heads can get by with 128 Kbps, but a high-quality call can require 500 to 768 Kbps.

Streaming-media technologies serve on-demand and live video for corporate communications, promotion, sales and training. How tightly the video codecs compress video data for transmission dictates the streaming bandwidth requirements. Codecs with tight compression algorithms, such as RealNetworks' RealMedia, use as little as 20 Kbps of bandwidth. But high-quality video and other corporate LAN communications generally use MPEG standards for MPEG-1 and MPEG-2. This quality ranges from 30 frames per second for full-motion video to 60 interlaced fields per second to accommodate broadcast video.

MPEG-1 can take up to 1.5 Mbps per stream. MPEG-2 generally requires 2 Mbps to 6 Mbps but can need as much as 40 Mbps. MPEG-4 calls for less than its predecessors and targets the low bandwidth requirements on the Internet along with Apple Computer, Microsoft and RealNetworks. It needs 64 Kbps to 4 Mbps.

Traffic Prioritization

Bandwidth alone won't ensure the smooth delivery of real-time voice and video packets, especially considering IP's best-effort nature over Ethernet. You'll likely need additional guarantees under heavy loads, so consider upgrading to Layer 3 switches and traffic prioritization strategies such as RSVP (Resource Reservation Protocol), DiffServ (Differentiated Services) and MPLS (Multiprotocol Label Switching).RSVP is designed to clear a path for audio and video traffic to reduce packet delay and loss. This flow-based communication protocol, which signals a switch or router to reserve bandwidth for a real-time transmission, such as a VoIP (voice over IP) call, allows a node to specify an end-to-end delay, such as 100 ms.

RSVP does not scale well, however, as every device along the path from packet origin to destination needs to maintain the state of the connection. Therefore, small to midsize enterprises can benefit from RSVP; large enterprises must discriminate services without maintaining a per-flow state at every switch. QoS strategies such MPLS and DiffServ will likely be more suitable.

U.S. PBX Vs. IP-PBX Line Shipments

click to enlarge

Implementing QoS with MPLS is like building a circuit-switched network that can create end-to-end circuits over any type of Layer 2 transport medium. In an MPLS network, a LER (Label Edge Router) assigns incoming packets a "label." Packets then travel along an LSP (Label Switch Path) that makes forwarding decisions based on the label's contents. At each hop along the path, an LSR (Label Switch Router) strips off the existing label and applies one with new forwarding instructions. Network operators establish LSPs to route around network congestion or create IP tunnels for network-based VPNs.

MPLS standards are still advancing through the IETF, and implementations have evolved differently, depending on the vendor. For example, Cisco Systems implements MPLS as Tag Switching. If you standardize on routers and switches from Cisco or another vendor that supports MPLS, your voice and video traffic will get priority without RSVP's heavy overhead. In addition, some private network providers, such as Masergy Communications, employ MPLS to move converged traffic to your remote offices.

If your enterprise is not a Cisco or Extreme Networks shop and you have heterogeneous networking equipment, you can use a QoS option such as DiffServ. Layer 3 switches (ASIC-based routers) can now perform route lookups at sufficient speeds to obviate label switching. Rather than tag traffic the way MPLS does, DiffServ modifies bits in the IP header to indicate QoS. DiffServ-compliant switches, such as Extreme Networks' Summit and BlackDiamond devices, read the value stored in the IPv4 packet header's ToS (Type of Service) or IPv6 packet's Traffic Class octet.

The value or DSCP (Differentiated Services Code Point) is 6 bits wide and capable of 64 classifications. The default code point (000000) represents a common IP packet and maintains IP's best effort in forwarding. Code points for preferential treatment (11x000) are given priority queuing and can be forwarded using a variety of mechanisms (such as strict priority queuing, weighted fair queuing and class-based queuing).

Although DiffServ may appear to be the QoS of choice for large enterprises, you need to test this solution in your environment. For example, vendors may use different forwarding mechanisms for the same DSCP. Other environments with Gigabit Ethernet backbones may have little need for QoS on the LAN, but may need to prioritize traffic over a saturated T1 WAN link. In that case, both a QoS strategy and a packet shaper, such as Packeteer's PacketShaper, may provide the requisite assurances that voice and video traffic will receive priority throughout the enterprise.Tools such as NetIQ Corp.'s Vivinet Assessor can analyze enterprise environments and traffic patterns and assess the impact of voice traffic, using codecs such as G.711. In the future, such tools will evaluate the impact of video traffic. Taking the QoS route is only one way enterprises can actively manage their IP traffic. Content delivery networks and multicasting can also reduce the bandwidth requirements for convergence. These technologies are critical for enterprises that reuse content both in-house and over the Internet to generate revenue, maintain content supply lines to a remote workforce, and reduce the costs in corporate communications.

As Web pages grow fat with embedded audio and video content, network pipes to deliver that content have starved. Most enterprises are still using T1 lines as connecting points for WANs and the Internet. These points can create bottlenecks and congestion as customers and employees request content from a central location. For example, 10 branch employees accessing a streamed training video encoded at 150 Kbps from a central server can eat up the bandwidth of a T1 line and saturate the link to your branch office. Likewise, high loads from customers can make your public Web site unreachable.

By positioning content near the network's edge, public CDNs (content-delivery networks), such as those offered by Akamai Technologies and Speedera Networks, can reduce the latency in content delivery by leveraging cache. Using cache devices or servers strategically located in ISP PoPs (points of presence), public CDNs mirror or preposition original content from the enterprise and make it available close to customers and end users. The result is more accessible content and faster delivery. In the enterprise, eCDNs (enterprise CDNs), such as Volera's Velocity CDN and its Excelerator server cache, provide similar functions.

Enterprise cache servers are positioned close to original content servers or near end users on remote WAN links to speed content to end users. The servers dynamically pull content from origin servers based on end-user requests and maintain that content in cache for later use. Cache servers (also called proxy caches, since they act on the origin servers' behalf) also store prepositioned content copied or mirrored from central servers. Prepositioning reduces the amount of information that traverses costly links, improves application response time and reduces overall network latency.

Proxy caches can also integrate with directory and authentication services to provide access control to cache servers and content. They support a variety of file types and content from FTP and streaming media servers from Apple, Microsoft Windows Media Technology and RealNetworks. Caches can serve streaming media on demand or live, and limit the bandwidth for narrow pipes. For live events, caches support stream splitting, in which a single streaming file is received over a WAN link and split over the LAN to downstream users. This one-to-many delivery mechanism leverages multicasting in the enterprise.Multicasting delivers a single stream from a source such as a streaming media server to a group of receivers or a multicast group. Rather than send separate streams to each user request, as in unicast, multicast streams serve many end users from one stream. This one-to-many or many-to-many communication technology efficiently disseminates information from a source to a set of receivers and reduces overall bandwidth requirements (see "The Wizardry of Multicast").

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].

Active Voice: The PC-based voice-processing provider has extended its unified-messaging solutions to Microsoft Exchange in Kinesis.

Apple: Apple's QuickTime 6 supports MPEG-4 video and audio file formats. And the company has recently acquired Emagic, Nothing Real, and Prismo for digital audio and video creation.

Blue Coat Systems: Formerly CacheFlow, Blue Coat Systems refocuses cache servers as security gateway appliances in its SG800 and SG6000 devices.

Interwoven: The company's Media Asset Management platform for content delivery extends its Enterprise Content Management suite.LVL7: The company's Fastpath 3.0 software for network processors and ASIC/ASSP includes advanced services for QoS, multicast and BGP-4.

Microsoft: Windows Media 9 Series supports MPEG-4 video (not audio) file formats.

MPEG-4 LA: MPEG-4 patent holders, including Microsoft, Philips and Sony, are pursuing license strategies for MPEG-4 video on per-stream and decoder schemes.

NEC: NEAX Internet Protocol Server Distributed Model (IPS-DM) extends NEC telphony solutions over IP.

NetIQ: The company's Vivinet family of products to assess and manage VoIP and IP telephony solutions now includes diagnostic software for root-cause analysis.Radvision: Radvision has begun shipping Invision, a 1U videoconferencing appliance that includes an integrated gateway, gatekeeper and multipoint conferencing unit with ISDN and IP support.

RealNetworks: Helix Universal Server 9.0 and RealVideo 9 are fully interoperable with the MPEG-4 specification.

Sonexis: Sonexis' audioCollaborator is an IP audioconferencing system in a 1U appliance that supports up to 96 PSTN ports or 120 IP connections.

Sony: With Philips, Sony acquired InterTrust to develop a DRM (digital rights management) solution.DiffServ (Differentiated Services): IETF's architecture for scalable services differentiation in the Internet

H.230: ITU-T standard provides frame-synchronous control and indication signals for audiovisual systemsH.234: ITU-T standard for delivering encryption key management and authentication system for audiovisual services

H.323: ITU-T's packet-based multimedia communications system for IP telephony

ITU-T's terminal for low-bit-rate multimedia communication

H.324: Point-to-point protocol between cache servers and network-based applications

ICAP (Internet Content Adaptation Protocol): IETF's model to transport audio, video, real-time and traditional data within a single architectureMEGACO/H.248: IntServ (Internet Services): IETF and ITU-T proscribed protocol requirements for the MGCP between a Media Gateway Controller and a Media Gateway

MGCP (Media Gateway Control Protocol): IETF API and protocol for controlling VoIP gateways from external calls

MPEG-1, -2, -4: Standards for the compression of digital audio and video

MPLS (Multiprotocol Label Switching): IETF solution for integrating label-swapping QoS framework with network routing

RSVP (Resource Reservation Protocol): IETF protocol for Internet servicesRTP (Realtime Transport Protocol): IETF communication protocol for real-time transport of audio and video over IP

SIP (Session Initiation Protocol): IETF's text-based protocol to initiate interactive communications between users

T.120: ITU-T's data protocol for multimedia conferencing

WCCP v.2: Cisco's communication protocol for cache serversEven as enterprises adopt gigabit per second switch technology, T1 leased lines--a technology holdover from the 1960s--continue to be the most common bandwidth enterprises use to connect LANs and tap the Internet. T1 lines carry both voice and data over 24 channels of 64 Kbps with TDM (time-division multiplexing).

Enterprises that need more bandwidth than T1's 1.54 Mbps can install additional T1 lines or lease fractional T1s in two-channel, 128-Kbps increments. But at approximately $800 per Mbps, speed can kill. There has been movement from private leased lines to less expensive alternatives in shared networks, such as frame relay and ATM. But frame relay is a poor choice for voice traffic because it lacks the guaranteed quality of service found in ATM networks.ATM remains the only way to deliver data, video and voice over one network with proven QoS (quality of service). As the demand for converged network applications increases, the U.S. market for this technology is projected to grow from $4.7 billion in 2001 to $9.9 billion in 2005, according to the 2002 Telecommunications Market Review and Forecast.

By breaking down information into 53-byte cells, ATM gives voice traffic a high priority without forcing it to wait behind long data packets, the way Ethernet does. A 5-byte header atop a 48-byte payload establishes priority, routing information, packet-sequencing and error-checking. Unlike Ethernet, ATM is switched in hardware without table lookups and therefore reduces latency, another crucial factor in dealing with time-sensitive video and voice traffic. Various media types, such as twisted pair, coaxial and fiber optic cable, support ATM, which scales from 56 Kbps to OC-12 (622 Mbps).

Because installing ATM is more expensive and complex than installing Ethernet, ATM is most commonly used in carrier backbone networks. But 12 percent of the U.S. ATM equipment market consisted of enterprise WAN switches in 2001 (2002 Telecommunications Market Review and Forecast). Enterprises use ATM for multimedia transmissions and videoconferencing. For example, ATM interconnects large health care organizations to transmit X-rays and images and to engage in videoconferencing.• RFP/RFQ Builder

White Papers

Research ReportsCareers




Stay informed! Sign up to get expert advice and insight delivered direct to your inbox

You May Also Like

More Insights