...voice is free…

CONTENTS

1. Introduction

2. Background

3. Technologies

   1. Frame Relay

   2. IP

   3. Asynchronous Transfer Mode (ATM)

4. Standards

5. Potential Applications

6. The Marketplace

7. Implementation Issues

   1. Physical Connectivity

   2. Network Design Considerations

   3. Other Technical Considerations

   4. Intranet versus Internet

8. Summary, Conclusions

VOICE OVER
[Presented by Peter Lupica and copyrighted on March 10, 2000]

1) INTRODUCTION

...voice is free... This seems to be a popular phrase these days and is usually used in reference to a so-called, "converged" voice and data network. The rationale states that a data communications infrastructure is likely already in place to serve the needs of the corporate data environment. Some of this network capacity may be going unused. If this unused bandwidth is to support voice traffic, or other similar types of traffic such as video and fax, then greater network utilization efficiencies result. In effect, voice calls, (or FAX or video) may be able to ride for free on an existing data network. Therefore, "…voice is free…".

How does this work and is it really free? Those are valid questions and while this Paper may not provide all of the answers it will attempt to shed some light on the matter and provide enough background information to ask the right questions.

2) BACKGROUND

Since the advent of digital PBX's and digital carrier systems - T1 and the rest of the hierarchy - voice has been digitized. If one were to compare "voice bits" with "data bits", you would quickly realize that there is no discernible difference. However, the characteristics of each type of service tend to be significantly different with voice, FAX and video tending to have similar characteristics.

These differences tend to revolve around the nature of each type of communications, the resultant network tailored to effectively support the nuances and user levels of expectation. It may be helpful to at first consider some of the general characteristics such as those outlined below.

Data
 

• "Calling" pattern may be more predictable if not controllable
   

• Per-packet addressing is required to assure data delivery, with no requirement for "call" setup.

• Relatively time insensitive with regard to successive segments or portions of the "call";

• Network resources are consumed as needed or on demand and typically not reserved on an end-to-end basis;

• Users may be more tolerant of disruptions that effect data transmission. They have all experienced difficulties with their own workstations, application programs, etc. and may be (slightly) more understanding or tolerant.

Voice, Fax, and Video

• Intermittent calling pattern in that calls are initiated at the whim of the user;

• Call setup is required to initiate the connection;

• Time sensitivity with respect to successive segments or portions of the call in that they need to "arrive" within a specific window;

• Network resources are reserved for the duration of the call;

• Fixed, dedicated bandwidth (typically 64 Kbps per voice channel); 

• User experience is based on the U.S. telephone network and its' reputation for highly reliable service. Call setups occur in a matter of a few seconds, calls almost always completed to the destination, and mid-call disconnects rarely occur.

Also, with regard to voice, there is a characteristic of human speech that is of importance. That is the fact that it contains a tremendous amount of redundant information. In general, only about 22 percent of a typical dialog consists of essential speech components that need to be transmitted for complete voice clarity with the balance, 78 percent, being made up of pauses, background noise, and repetitive patterns. The removal of silent periods and redundant information through advanced techniques enables voice to be efficiently "compressed". Then when the voice bits are packetized, the packets or frames (both terms are often used interchangeably) tend to be smaller than average data frames. As a result, lower bit rates can be effectively used.

However, unlike most data communications, which can tolerate delay, voice communications must be performed in near real time. This means that transmission and network delays must be kept small enough to remain imperceptible to the user.

3) TECHNOLOGIES

In theory, any data network protocol could support digitized voice. Voice over Ethernet has been tested in the lab for several years while Asynchronous Transfer Mode (ATM) was designed specifically to handle multi-media traffic, i.e. voice, data and video. Today however, most of the attention is focused on Frame Relay and/or IP networks.

a. Frame Relay

Frame Relay is a Layer 2 protocol and as such does not guarantee end-to-end frame/packet delivery. There is no specific provision to support the notion of "priority" traffic and congestion control is achieved by discarding frames/packets. In the event this would occur, it is incumbent upon the terminal equipment to recognize this and retransmit frames/packets accordingly.

Frame Relay has gained acceptance as a means to provide end users with a solution for LAN-to-LAN connections and other data connectivity requirements. Besides providing a flexible and efficient data transport mechanism, frame relay lowered the cost of bandwidth for tying together multi-protocol networks and devices.

Now there is considerable interest in Voice over Frame Relay (VoFR). The basic premise is that it has the potential to provide end users with greater efficiencies in the use of access bandwidth by functionally integrating voice, data, and fax over a single access link. This would result in offering a cost effective option for voice traffic transport needs between company locations.

In addition to providing basic services such as encapsulating data traffic for transport over the frame relay network, voice capable FRADs (Frame Relay Access Device) may sometimes provide connectivity between PBXs and other voice equipment.

Although work on Implementation Agreements for carrying voice over frame relay is progressing within organizations such as the Frame Relay Forum's Technical Committee, there is currently not a uniform standard or implementation agreement defined for vendor equipment interoperability or for the transport of voice across a carrier's public frame relay network. In the absence of such an agreement, many equipment vendors have developed proprietary methods for integrating voice onto frame relay networks. But even in the absence of full standards, there are end users who have successfully deployed voice over their frame relay networks.

b. IP (Internet Protocol)

IP is commonly used in a large number of today's data networks. It is a Layer 3 protocol and is primarily connectionless, i.e. no dedicated end-to-end path or connection. With regard to packet delivery, it is a "best effort" protocol. Again, it is incumbent on the terminal devices to manage and control packet delivery, re-transmitting data as necessary.

Voice over IP (VoIP) is a method of taking digitized voice and dividing it into packets. These voice packets are then sent with other packets (data, e-mail, video images) across a packet-switched network. At the receiving end, the voice packets are re-assembled packets and hopefully arrive as a normal sounding voice call. In some respects IP packet switching is more "efficient" than circuit-switching in that it only occupies the network as traffic requires bandwidth rather than tying up an entire circuit.

Because voice is packetized, it can be particularly challenging. Individual packets can potentially take different routes and arrive at the destination at different times. In a text message this can be reconstructed once all the packetized pieces have arrived. But in "real-time" voice communication, timely delivery to avoid significant pauses is required.

However, latency is not the only troublesome issue that must be dealt with. In addition there are the issues of dropouts (lost packets), security, numbering and perhaps most importantly, the lack of established Quality of Service (QoS) standards.

Advances in router and switching technology are aimed at resolving many of these issues and work is being done on the standards front as well.

c. Asynchronous Transfer Mode (ATM)

ATM is a packet or switching technology. It is based on handling fixed-size cells with each cell composed of "overhead" bytes as well as information or payload bytes. Among other things, the overhead data is used to give some indication as to the type of information that is contained in the payload.

The ATM scheme was developed specifically to effectively handle multi-media traffic – voice, data and video. ATM is aimed at using relatively high-speed links such as DS3, OC3 and higher, and in general is more costly to deploy than either Frame Relay or IP. As a result, ATM is currently employed as primarily a backbone technology, in very large user networks and in carrier networks.

The relationship between ATM and the other packet technologies described above seems to be evolving to one of co-existence. The current thinking is that VoFR and VoIP networks can/will be interfaced to ATM networks in order to provide for high speed, long haul transport.

4) STANDARDS

These days, any discussion about telecommunications must include something about standards. Due to the recent attention of using IP and/or Frame Relay networks for voice, fax and video, a number of organizations, including recognized governmental bodies as well as ad hoc committees, are involved in the standards process. Some of them are:

• American National Standards Institute (ANSI)

• Institute of Electrical and Electronics Engineers (IEEE)

• International Organization for Standardization (ISO)

• International Telecommunication Union (ITU)

• Internet Engineering Task Force (IETF)

• Electronics Industries Association (EIA)

• The Frame Relay Forum

• The ATM Forum

• The Voice-over-IP Forum

• Telecommunications & Internet Protocol Harmonization Over Networks (TIPHON)

Voice, fax and video over Frame Relay and/or IP networks touch both the voice and data aspects of communication technologies, and therefore the standards must address both of these elements. A sampling of some of the key and relevant standards include:

• The H.320 Series published by the International Telecommunication Union - Telecommunication Standards Sector (ITU-T), defines video telephony communications over point-to-point and multi-point connections. The most notable standard in this series is H.323, which specifies packet-based multimedia systems. The Internet telephony industry has adopted H.323 as the software standard for interoperability.
   

• The ITU "Q" series which addresses signaling, among other things.
   

• The G.700 Series published by the International Telecommunication Union – Telecommunication Standards Sector (ITU-T), defines the technical aspects of terminal equipment, including the specific algorithms that are used to encode and decode analog signals (such as voice) for transmission over a digital network. An example of a standard in this series is G.723.1, which defines voice encoding at 5.3 and 6.3 Kbps.
   

• Compression algorithm standards such as Microsoft's G.723 and Micom's G.729.
   

• The Request for Comments Series published by the Internet Engineering Task Force, specifies the operation of the global Internet and are used by many corporate intranets that connect to the Internet. Examples of documents that are used in voice, fax and video over IP implementations include RFC 1889 (Real Time Protocol), RFC 2068 (Hypertext Transfer Protocol), RFC 2205 (Resource Reservation Protocol), RFC 2326 (Real-time Streaming Protocol) and RFC 2327 (Session Description Protocol).
   

• The VoIP Protocol Stack published by the Voice over IP Forum defines the specific ITU-T and IETF standards that are required in support of a multi-vendor, interoperable protocol implementation.
   

• DIFFSERV (Differentiated Service) is a means of establishing some sense of "priority" for certain types of packets, i.e. voice. This is being pursued since it is felt that RFC2205 (Resource Reservation Protocol) will have difficulty scaling.

As the saying goes... "the nice thing about standards is that there are so many from which to choose". It bears repeating that the above is merely a small sample of standards, which may be appropriate in this environment. Understanding the importance of the standards, and knowing their key elements, is crucial if one is to attempt to implement a converged network.

5) POTENTIAL APPLICATIONS

To paraphrase a popular commercial of some years ago …."where's the beef?….". In other words, aside from the glitter of being on the very cutting edge of technologies, what is this good for? Here are some possible applications.

• Customer Support Call Centers: One of the fastest-growing segments is web-enabled call centers. Call centers are encouraging web surfing customers to push a "click and talk" button on their web sites rather than log off and wait for a call back. Instead, the "caller" simply clicks a button and instantly connects to an agent. MCI's Vault architecture lets customers on a web site simultaneously talk to a customer representative. Sprint also has a service "give me a call" that lets a Web surfer place a voice call from a browser.
   

• Fax: Fax currently accounts for 40% of all traffic on traditional voice networks translating to over $30 billion per year for just sending and receiving a fax. Users get the same cost saving benefits on fax transmissions as on voice calls.
   

• PC to PC: This is possible when both PCs are equipped with a microphone and speaker. The user "dials" the phone number of the person they want to reach, and the person on the other end can respond.
   

• Video Conferencing: The addition of a video camera will provide video conferencing capability as well as connect to Microsoft's NetMeeting application. There is a feature for white boarding that allows conference participants to review and edit documents in real time.
   

• PC to Telephone or Telephone to PC: Users can place and receive calls to any ordinary telephone. This is the likely set up for a web surfer connecting to a call center but it may also be the set up for someone at a PC on an office LAN. A properly equipped device places a call. The digitized voice packets are sent over the packet network. At the distant end, the call hops off the network at a gateway. The voice data is reassembled, decompressed and converted to the Public Switched Telephone Network (PSTN) format. It is then sent over PSTN to its destination. In the other direction, a remote worker communicating with the home office may place a telephone call that reaches a gateway. The data is digitized and compressed. The call is then routed over the packet network and finds the PC on the other end by using the unique IP address.
   

• Telephone to Telephone: The digitized voice packets are sent over the packet network via a gateway. At the end point the data hits another gateway, converted back to PSTN format and sent over the PSTN to its destination.
   

• It's also an efficient connection for a person retrieving voice mail and returning voice messages via laptops in remote locations such as hotel rooms. This combines the "low" cost of packet network connections with the ubiquity of telephones on the PSTN.

6) THE MARKETPLACE

At the end of 1995, the estimated number of active packet network telephony users was 500,000, which represented $20 million in revenue and 600 million call minutes. The market now is estimated at over 16 million users, translating to a $560 million business. Probe Research predicts that packet-based networks will carry more than 7% to 11% of the world's international phone traffic by 2002, 34% of the U.S. domestic long distance traffic, and 10% of the world's fax communication.

According to a Frost & Sullivan study released in 1997, the total equipment market for packet switched voice products is estimated to be over $2 billion by 2001 and $16 billion by 2004. These will be primarily telephony gateway platforms that bridge packet networks with the public switched telephone network. to enable long distance calling from telephone to telephone, fax to fax, PC to telephone, telephone to PC and web browser to telephone.

As should be apparent, the scope of packetized voice has become International with one of the driving factors being the ubiquity of the Internet. Along with this is the focus on IP rather than Frame Relay with the latter being viewed as an "edge network" technology. While perhaps beyond the scope of this article, it is worthwhile to devote some attention to this phenomena since there should be some long-term strategic reason for an enterprise to attempt a converged network.

According to some experts, ISPs are the next generation of telephone companies who will provide voice-over-packet-network service. They even predict that in the future, the present public switched telephone network will migrate to voice-over-IP network and that "voice" will merely become another data application.

A market representing that kind of potential is sure to attract a large number of potential players, ranging from start-ups to some very large, formidable companies. For example:

• The following carriers are/have conducted Internet telephony trials; USA Global Link, IDT, Delta3, WorldCom, AT&T, MCI, U.S. West, Bell Atlantic, Sprint, AT&T in Japan, KDD in Japan, Dacom in Korea, Deutsche Telekom in Germany, France Telecom, Telecom Finland and New Zealand Telecom.
   

• IDT Corporation introduced a service called Net2Phone Direct which provides PC to phone service throughout the world, and phone to phone Internet telephony services in the U.S. Users of both services can call any telephone around the world using the Internet at rates 95% less than local phone companies. Calls from local numbers to anywhere in the U.S. cost 8 cents per minute any time, UK at 18 cents per minute, Australia at 20 cents per minute while MCI charges $1.42 a minute, and Japan at 29 cents per minute.
   

• ITXC is an example of a new entrant Internet carrier. Its WWexchange service is designed to connect member Internet telephony service providers. ITXC is a start up company backed by AT&T and VocalTec and will offer network infrastructure and management for Internet telephony service providers so they don't have to create their own infrastructure, billing and administration.
   

• Bell Canada formed a separate division called Emergis designed to cannibalize the parent business by creating an advanced network based on the Internet to carry an increasing proportion of the parent company's voice traffic.
   

• Level 3 is spending billions of dollars to build a new IP network. It will create the first business-focused, pure Internet based local and long distance carrier with a new economic and technology model. The goal is to deliver services at 1/27th the cost of today's traditional circuit-switched networks. It will integrate voice, data and video over high bandwidth transmission facilities connecting an IP network architecture. Level 3 plans to take advantage of a ground-up IP network design to deliver services instead of converging circuit-switched and IP-based networks like most incumbent carriers. Level 3's approach running all traffic over a single network is easier, cheaper and quicker to manage and upgrade.
   

• WorldCom announced that it would introduce a worldwide Internet telephony service.
   

• America Online has introduced a voice over the Internet service for 9 cents a minute. And, connecting to www.internetcollect.com  will allow users to make 8 cents per minute collect calls anywhere over the Internet.
   

• Deutsche Telekom and VocalTec made a joint announcement in which Deutsche Telekom will purchase more than $30 million of VocalTec products, services and support for a planned rollout of Internet telephony services. Deutsche Telekom will become the first major carrier to develop a portfolio of Internet telephony services to supplement its existing telephony services worldwide. The two companies are already working on the "T-NetCall" pilot in which Deutsche Telekom is giving 1,000 selected customers the ability to communicate via the Internet using a conventional or mobile telephone.
   

• Bellcore has formed a new business called Soliant Internet Systems, it is intended to make Internet telephony as easy as using the public switched network.
   

• Bell Labs has formed a new division called Elemedia to bring Internet components to the marketplace. One of its first developments, voice compression software, is used by Lucent Technologies in the Internet Telephony Server. This server-based solution running on Compaq Proliant computers places voice and fax calls over the Internet. Elemedia has also licensed its voice coder software to firms who are implementing products that transmit voice over IP networks.
   

• Hardware vendors such as Nortel and Cisco Systems have announced, and in some cases can deliver, "voice-over" products.

While the convergence of voice and data has been predicted as being eminent for the last twenty years or so, it does appear that it is now closer to becoming a reality. However additional work must be done. Key challenges for the suppliers of voice-over technology include continued improvement in voice quality, adoption of interoperability standards and a reduction in the cost per port. A target cost of $500 per-port is usually required before significant deployment can be feasible. Current solutions range upwards from $1,500 per port and may require significant hardware and software changes/upgrades to embedded systems, both voice and data.

7) IMPLEMENTATION ISSUES

As one would imagine, there are several aspects to implementing packetized voice networks. From a technical perspective, differences among the various technologies - Frame Relay, IP and ATM - can be rather significant. The balance of this article will address some of the implementation issues which must be addressed, however ,the primary focus will be on Frame Relay and IP since these are the technologies which are currently drawing the most attention in the area of converged networks.

a. Physical Connectivity

The following drawing depicts the physical connectivity from a voice device to a packet network. From a high level, there is very little difference between the connectivity required for VoFR and VoIP. Basically what is required is a connection from the voice device to an interface device and then to the packet network. The connections from the voice devices may be individual trunks, T1 or in some cases, 10/100BaseT.

Frame Relay networks require a device known as a FRAD (Frame Relay Access Device) where with IP networks, access can be accomplished either directly to a router or to an intermediate hub and then to the router and finally to the network. Access links, from either the FRAD or the router, can be any speed with Frame Relay being generally limited to T1 as a maximum.

b. Network Design Considerations

Careful attention to the design of the network is of paramount importance if one is to attempt to converge services over a single network whether the network is already in place or being implemented new.

In addition to the various standards and protocols, one must understand the nature of the packet technologies involved. Several examples are outlined below.

• Frame Relay uses a mechanism called the Committed Information Rate (CIR). CIR's can be used as an attempt to control traffic, however in the carrier world CIR's are usually tied to cost – the higher the CIR the higher the cost. Each link is assigned a CIR and it is at this rate that reliable communications are supported. They are usually established well below the speed of the link since traffic can "burst" above the CIR. However the packets that are in excess of the CIR are marked as being "Discard Eligible". During periods of network congestion, these packets will be the first to be discarded, with no advice given to the originator. With regard to data, this results in little more than a nuisance since the terminal equipment takes care of any necessary re-transmissions to insure that the distant end receives a complete message. With voice however, it becomes more of an annoyance since the re-transmission mechanism found in data terminal equipment is generally not present in voice equipment.
   

• Frame Relay networks currently can only support Permanent Virtual Circuits (PVC) meaning that communication can only take place between the two designated end points. The concept of tandem switching would be somewhat impractical to implement at this time.
   

• IP is a routed protocol and as such "switching" capability is inherent merely by addressing packets to a different destination address. However, routing is accomplished on a per-packet basis and as a result successive packets of the same message could traverse different routes/paths through the network. The result could result in excessive jitter (see discussion below).

Other important design considerations, which warrant close attention, include:

• Overall bandwidth requirements for all service types;

• Quality of Service (QoS) targets/standards;

• The transmission speed of the network infrastructure;

• Physical topology and physical media of the network infrastructure;

• Minimum and maximum packet sizes;

• Traffic measurement and engineering

• Control functions that are required to set up and maintain the connection.

• Jitter, the variation in arrival times between packets, may require the incoming packets to be placed in a buffer and then released from that buffer at standard intervals.

• Latency, or the delay from the signal source to the signal destination through the network. This is a key element effecting quality of service. With most systems, a round-trip latency of 400 milliseconds is considered the maximum tolerable delay, with a round-trip latency of 200 milliseconds more optimal. When latency exceeds these limits, the quality of the received voice signal degrades dramatically.

c. Other Technical Considerations

In addition to the above, there are several other technical aspects, which must at least be considered.

Voice Compression:

Uncompressed digitized voice and fax require a large amount of bandwidth. This often makes it impractical to transmit these signals over low-speed access links. The use of low bit rate voice compression algorithms can make it possible to provide high quality speech while using bandwidth efficiently. Various algorithms are used to sample the speech pattern and reduce the information sent while retaining the highest voice quality level possible. The general function of these strategies is to scrutinize the speech signal more carefully, to eliminate the redundancies in the signal more completely, and to use the available bits to code the non-redundant parts of the signal in an efficient manner. As the available bit rate is reduced from 64 Kbps to 32, 16, 8, and 4 Kbps or below, the strategies for redundancy removal and bit allocation need to be ever more sophisticated.

Echo Cancellation:

Echo occurs when the transmitted voice is reflected back to the point from which it was transmitted and becomes more noticeable as the propagation delay increases. The longer the distance, the more delay, and the more likely echo will result. The current voice network employs echo suppressors to overcome this phenomenon however data/packet switched networks do not use echo cancellation equipment in the network therefore it is up to the equipment vendor to address echo cancellation in the equipment.

Delay and Delay Variation:

The bursty nature and variable frame sizes of packet networks may result in variable delays between consecutive packets. The variation in the time difference between each arriving packet is called "jitter". Jitter can impede the ability of the receiving end customer premise equipment to smoothly regenerate voice. Since voice is inherently a continuous wave form, a large gap between the regenerated voice packets will result in a distorted sound.

Frame/Packet Loss:

Packetized voice can usually withstand infrequent packet loss. If a voice frame/packet is lost, the user will most likely not notice. If excessive frame/packet loss occurs, it is equally unacceptable for voice as well as for data traffic however in the case of voice, it will likely be much more annoying to the users.

Traffic Integration

There appears to be some mimicking of the switched public voice services network with regard to supporting fax and data modem services. This ability may prove to be beneficial to end users that have high fax traffic volumes between locations. However, it is difficult to reliably compress fax and data modem signals to achieve the low bandwidth utilization often necessary for the most efficient integration. Some implementation schemes compress voice to a low bit rate, but upon detection of a fax tone, the bandwidth is reallocated to a higher bit rate to allow for faster fax transmission.

Prioritization:

Voice, fax and some data types are delay sensitive. This means that if the end-to-end delay, or delay variation exceeds a specified limit, the service level will degrade. To minimize voice traffic delay, a prioritization mechanism that provides service to the delay sensitive traffic first can be employed. However Vendors may choose to use a variety of proprietary mechanisms to ensure a balance between voice and data transmission needs.

Fragmentation

Fragmentation is a technique used to break up larger blocks of data into smaller, less delay-creating frames. This is another means used to ensure the highest voice quality level possible. Fragmentation attempts to ensure an even flow of voice frames into the network, minimizing delay. The fragmentation often involves all of the data in the network to retain consistent voice quality. This is because even if the voice information is fragmented, delay will still occur if a voice frame is held up in the "middle" of the network behind a large data frame. Additionally, fragmentation reduces jitter because voice packets can be sent and received more regularly.

Digital Speech Interpolation:

Digital speech interpolation addresses silence suppression. The nature of speech communication entails pauses between words and sentences. Advanced voice compression algorithms, which identify and remove these redundant patterns, effectively reduce the amount of speech information to be transmitted.

Multiplexing Techniques

Some equipment vendors use different bandwidth optimization multiplexing techniques such as Logical Link Multiplexing and Sub-channel Multiplexing. Logical Link Multiplexing allows voice and data frames to share the same PVC (Permanent Virtual Circuit). This can provide savings on carrier PVC charges and it increases the utilization of the PVC.

Sub-channel Multiplexing is a technique used to combine multiple voice conversations within the same frame. By allowing multiple voice payloads to be sent in a single frame, packet overhead is reduced. This may offer increased performance on low speed links.

Intranet versus Internet

Another primary consideration has to do with the scope of the converged Network. Will it be confined to the Corporate Intranet only or will use of/access to the Internet also be allowed?

It is likely that Intranet-only deployments will enjoy more success than those making use of the Internet. The reasons being that the Corporate Intranet can be closely monitored, controlled and tuned so as to provide acceptable levels of service for all types of communication. Bandwidth can be added and/or re-arranged, routing can be controlled and priority schemes can be implemented even though they may be vendor proprietary. In other words, an organization has, and can exercise complete control over the environment.

The Internet on the other hand, is a rather loose confederation of "service" providers, each with their own goals, objectives and areas of interest. A single organization has little if any control over the environment. Consider your own experiences when trying to navigate the Internet during peak times. Then consider how the delay you experience in downloading a file or waiting for a screen to load would impact a voice conversation. However, the allure of low/no cost service can be difficult to ignore.

 

8) SUMMARY / CONCLUSIONS

If you are going to attempt to change even part of the voice communication infrastructure from a circuit-switched to a packet-switched environment, you may be met with some fundamental technical challenges. First and foremost you must assure that the quality of service of the new system at least matches the quality of the old system. Delivering voice, fax and video signals from one point to another cannot be considered successful unless the quality of the delivered signal satisfies the recipient. Otherwise, you may have some very unhappy constituents on your hands.

If you move application traffic from an ultra-reliable network (such as the Public Switched Telephone Network) to a less-than-ultra-reliable network (such as the global Internet or your corporate Intranet), you should consider the following questions:

• What level of reliability can be achieved (i.e. five "9s")?

• What level of availability can be achieved?

• How will the converged network be managed?

• Will existing management policies/procedures be effective?

• Will the current staff be able to cope with the new environment?

• Will Service Level Agreements (SLAs) have to be restructured?

• What will be the impact of the loss of the quality commonly associated with toll traffic due to voice compression?

• What will be the impact of the loss of management and administrative benefits associated with carrier voice services (i.e. the loss of consolidated voice billing and invoice itemization, end user charge back capabilities, ID and accounting codes?

• How will the lack of equipment interoperability between customer premise equipment vendors be handled?

• What will be the impact of the lack of standards defining the acceptable levels of quality for voice transport over a carriers network?

In addition, there are other forces, some that are beyond your control, that will shape a migration to a converged network, especially if consideration is being given to using the Internet or some other public packet network.

• Will the IP Telephony Service Providers (ITSPs) be subject to federal regulations?

• Will access charges be imposed on VoIP calls?

• Does the Internet have the capacity for the expected increase in IP traffic?

• Will the key PBX vendors emerge as the strongest gateway suppliers, or will the router vendors mount a significant challenge from the data networking side and increase their market share in this new area?

• Are network management and protocol analysis vendors prepared to support this new market?

• How will carriers guarantee performance in the absence of standards?

• Are the carriers equipped to maintain and troubleshoot the new environment?

• How will the carriers deal with the fact that voice packetization occurs in equipment on the end users premise and outside of their network?

The above not withstanding, packetized voice, whether on Frame Relay or IP is here to stay. It is gaining in acceptance and the technology advances are beginning to address and resolve many of the inherent problems and issues. In some instances it can represent rather significant cost savings, to say nothing of the advantages which could be realized by truly integrated applications such as call center management.

As is usually the case, the road to successfully implementing new technologies, especially ones that hold the promise of so much value and benefit, is not without pitfalls that could be potentially serious. While some may consider the following as a "keen sense for the obvious", it is worthwhile to review some points which could be helpful if one is to investigate and/or pursue the implementation of packet switched voice.

1. Thoroughly research the technologies. Use the Web as well as information that can be obtained from prominent equipment vendors. However be aware that the equipment vendors have a proprietary interest.
    

2. Review business goals/objectives/strategies to insure that there is some relationship between them and any potential/perceived benefits of a converged network.
    

3. Put together a plan as to how to approach the implementation. Include the critical success factors that should be achieved.
    

4. IS and Telecom must work closely together, even if they are not in the same organization.
    

5. Inform management and users as to what is taking place. Focus on the business objectives and the benefits to each. Also be sure to include the possibility that the actual implementation may not be transparent to them. Keep them informed as to progress along the way.
    

6. Use outside help. Sources would include counterparts in other organizations, vendors (bearing in mind that they would have a proprietary interest) and others. Preferably, people who have had some experience with converged networks or at the very least demonstrate a thorough knowledge of the technologies.
    

7. Insure vendor interoperability. Consider conducting trials/tests during off hours. Use a documented test plan and record the results for later review and analysis.
    

8. Start small and use a phased approach. For instance, choose to integrate a few voice channels and/or serial data links over an existing connection between two or more company locations. Carefully monitor the results. Talk to the affected to users to find out about their experiences. Only after a successful real-world trial, proceed with integrating additional channels. Monitor the results at each step along the way.