Yet few IT shops have yet embraced voice/IP with open arms.
But for those wise enough to recognise a technology which is poised to come of age - particularly in areas such as call centres - there are a number of ways to get voice/IP up and running without dramatic changes to existing infrastructure.
Voice/IP is not expensive, but the initial outlay will depend largely on where and to what extent the technology is deployed. Some sites only want voice/IP for local telephony with all externals calls still going over the public switched telephone network (PSTN) or over a conventional private voice network using leased lines. But more commonly, voice/IP is deployed externally with the existing on-site telephone infrastructure left intact. In this case, a gateway is needed to route a call over a wide area IP network.
The biggest technical question is where and how best to deploy the gateway function. The best option will depend on factors such as the size of the site (see box on page 40). While there are serious questions over voice/IP reliability and quality, analyst GartnerGroup predicts that these are likely to fade this year.
Mike Hafferty, managing director of Vegastream - a specialist maker of hardware for IP telephony - argues that IP will ultimately prevail not because it is cheaper, but because it will offer better quality than the PSTN.
'Quality will be the main driver,' says Hafferty. 'In a year or two, we won't want to go back to the PSTN because it will seem so bad.' Current telephony systems are constrained by the bandwidth limit imposed by the PSTN. With the cost of bandwidth falling, and with voice accounting for a rapidly diminishing proportion of overall global transmission capacity, future IP networks can allocate bigger channels to each voice call while still cutting prices. Users may even be able to choose the quality of their calls and pay accordingly.
But this is the future. At present, the quality of voice/IP almost never exceeds the PSTN and suffers from variability, even over private IP networks.
Technology and standards to provide an end-to-end connection and offer guaranteed bandwidth without interference from other network traffic are only just coming into place.
Despite limitations, a few large multinationals are considering voice/IP for communications between internal sites where some sacrifice in quality is deemed acceptable.
The Pepsi group, for example, plans to communicate with its franchise-holders in remote parts of the world where local telecoms services are either expensive or of poor quality. In this case, voice/IP may be no worse than the normal telephone service, and extending the company's corporate network, perhaps via satellite, would be too expensive.
Pepsi's planned deployment of voice/IP is potentially one of the largest by a single organisation. Anthony Roberts, the company's European and African IT manager, says the technology will cut the communication costs of its franchise holders by 30%. Roberts admits that voice quality could vary. 'Because it's an unmanaged service, it could deteriorate. It's a risk, but it's only a supplement to voice lines,' he said.
Voice/IP is being more widely deployed over private IP-based intranets, but generally only to save money by routing conventional calls from existing phones over the IP backbone. 'There is stronger demand for voice/IP carried in the backbone where the devices are ordinary telephones,' says Keith Malinson, European managing director at researcher the Yankee Group.
In this case - and when virtual private networks provided externally by service providers are used to carry IP-based voice - the quality is similar to GSM cellular, and so is usually adequate for internal calls. However, even then the quality may not be deemed reliable enough for talking to customers.
The sacrifice in quality is not worth making if the savings are only negligible. For this reason, voice/IP for toll bypass (see box on page 38) is largely confined to international calls. Take the case of London-based marketing agency Rainier. The company implemented voice/IP for toll bypass using its virtual private IP network for calls between its London office and offices in Boston and San Francisco. Company director Stephen Waddington says voice/IP has cut Rainier's transatlantic voice communications bill by 75% compared with BT's rates.
Quality is not an issue because voice/IP is roughly comparable with cellular GSM, which has become widely accepted for business conversations, says Waddington. 'We occasionally get a slight delay at the start,' he says.
This delay, about 15 to 30 seconds, is the time it takes to set up the fixed path through the end-to-end network, including the component provided by the ISP. In other toll-bypass applications, this delay may be shorter.
Rainier is not using voice/IP for communications with its customers because of this delay, and because of the slightly poorer quality compared with the PSTN. The company is not using voice/IP locally within the UK because the cost savings are not significant enough to compensate for the inferior quality.
One problem with using intranets or virtual private networks to carry voice is that although the quality is adequate most of the time, it can become unacceptable during peak periods. The ideal solution would be to route calls over the PSTN at such times, if only you could tell in advance that the quality of service available over the IP network was inadequate.
Networking and telecoms systems vendor Nortel Networks has developed a technique for monitoring the condition of an IP network for this situation.
The company's Meridien private branch exchanges (PBXs) can assess the likely quality of the IP network on an ongoing basis by transmitting test packets and measuring the transit delays.
While this delay is kept within acceptable bounds, the network is deemed acceptable for voice/IP. But if the delay falls outside the bounds, calls are re-routed over the PSTN.
Cost savings aside, the value of voice/IP is its potential to deliver new applications and features not so easy to support over the PSTN.
The most significant of these new applications will be for web-enabled call centres, where voice/IP will enable voice calls to be opened up within an existing Internet session. For example, between a user and a call-centre agent while making an online purchase.
If that user needs help or is unable to complete the purchase via the web site, a voice link could be established with a agent using voice/IP within the session, without requiring a separate line.
At present, some ecommerce sites have 'call me' buttons, but the call-backs run on a separate line over the PSTN. This is unhelpful for most consumers with one line already busy during an Internet session. Voice/IP solves this problem and provides an incentive for ISPs to make the technology available on their networks to their larger corporate customers.
'Call centres are going to be one of the major drivers for voice/IP in the business sector,' says John Matthews, a specialist consultant for IP telephony at analyst Ovum.
Voice/IP will also enrich existing telephony by making sophisticated call-handling features more readily accessible. 'Modern PBXs have for years come with a huge sophistication of features which are largely inaccessible because the LCD display on handsets is not a useful interface,' says Kurt Christopherson, UK business development manager at network supplier 3Com.
'When you do voice/IP, your telephone system is the same as the PC system, and you can use wizards. So it becomes more like it learning you rather than you learning it,' he adds.
This makes it far simpler to perform basic functions such as call divert, but also to develop new ones, like more reliable identification of the incoming caller than is possible using the calling line identifier (CLI) within PSTN or ISDN public services.
The existing CLI can only identify callers who are using one of their regular phones such as their office extension or mobile, and then only as long as your system has a directory cross referencing the number with the name. You can transmit user names with the call using IP signalling, so that the recipient can tell at once who the caller is. Although not all calls will use an IP phone or PC and many will come via the PSTN, at least there will be the CLI to fall back on for identification.
In any case, service providers may offer caller identification as a feature with voice/IP, transmitting the name of an incoming caller in the form of IP packets so that this can be displayed on a PC screen. Then the called party can decide whether to answer, ignore or route to voice mail.
IP signalling, perhaps combined with caller identification, can also be used to facilitate sophisticated call conferencing of the kind previously only available as a specialist service. Larger conferences usually require a mediator to control admissions to the conference and introduce new participants to the others. The use of IP signalling combined with the PC interface makes it easy to perform these functions.
Providing these new features will help a new generation of service providers compete against the big players in the telecommunications field, says Ian Roberts, managing director of Telecomputing Europe, a specialist provider of value-added IP-based services for telcos.
'Existing telcos will kill the new ones on price alone because they have the economies of scale,' says Roberts. 'Next-generation telcos will only be able to survive if they can provide new services.'
However, voice/IP is likely to have just as significant an impact within the local area network (Lan), although for rather different reasons.
Of course, voice/IP will have to be available on an end-to-end basis to deliver some of the features already mentioned, but in some cases it will only be deployed locally where the aim is to save money by running a single set of cables on a single network, instead of separate ones for voice and data. It will also make moves and changes far easier and cheaper to manage because the PC and phone can both be shifted by the same IP network reconfiguration. This alone would save large organisations a fortune.
3Com is about to launch a complete PBX which operates over an Ethernet Lan and provides up to 100 lines. At present, this connects externally to the standard public network, and uses tables to translate the numbers of incoming calls into the Ethernet Mac address of the destination phone or PC. The system also converts internal dialled numbers into the corresponding Mac address.
'We are working with Siemens on development of the voice capabilities, and in 12 to 24 months we'll have a voice/IP private automatic branch exchange,' says Christopherson.
3Com says it will then be able to offer end-to-end voice/IP. Even in this case, however, IP would not necessarily be used to transport the voice within the core of the public network. An asynchronous transfer mode backbone was used to carry the first public voice/IP call from Ireland to the UK, made last month by Irish prime minister Bertie Ahern to the president of the ISP Interoute.
Some still argue that voice/IP will not be widely - or at least quickly - adopted. PSTN will continue to give voice/IP a good run for its money for some time both on price and features, says BT information liberation manager Chris Gahan. 'People get carried away by new technology and fail to take on board the benefits of the existing environment.
'Voice quality over data networks has always been a challenge, and so voice/IP will start as a niche application. It will not overtake the tried and tested and competitive PSTN quickly,' Gahan argues.
Gahan's argument is backed by a controversial report by Forrester Research which argues that voice/data integration delivers at best marginal cost savings when the levels of complexity are taken into account. But the arguments of the report do not apply to tactical use of voice/IP for immediate cost savings on particular routes, as achieved by Rainier. It also does not take full account of potential new sources of revenue from applications, such as fully integrated call centres that require IP-based voice.
On the contrary, the consensus of industry pundits indicates that we will live in a predominantly IP-based world by 2005, so ignore it at your own company's peril.
VOICE/IP's MAIN APPLICATIONS
Internet telephony and voice/IP overlap but are not synonymous. The former is the use of the Internet for voice-based applications. The connections are usually made via a standard dial-up connection, which could also be via a voice/IP connection within an Net session. Conversely, voice/IP can be implemented over any form of IP network including local area networks (Lans) and corporate Intranets.
Eight types of voice/IP application can be identified:
- On-site IP telephony allowing internal calls to be made over IP-based Lans. The main motives are savings in cabling, network management and administration
- Corporate toll bypass where voice calls are made over internal IP networks or virtual private IP networks. The motive is to save money, particularly for international calls
- Voice over the Net between two multimedia-equipped PCs. The main incentive is to make very cheap calls at local rates anywhere in the world. However, it is not suitable for most business calls
- Fax over the Net. The cost is very cheap, and quality is less of an issue because communication is not real time and loss of bandwidth simply delays transmission
- IP-based public phone services. Carriers can cut costs by consolidating voice and data over single IP core networks. Carriers will then be able to deliver new features not possible over the PSTN such as advanced conferencing
- Call-centre IP telephony. This is a variant of voice/IP , but is a distinct application
- Voice messaging over the Internet. The Net can become a medium for unified messaging. Quality is not an issue because timing is not critical when sending messages in any medium
- Video-over-IP. This a field unto itself raising a variety of unique technical issues.
ARE YOU READY FOR VOICE/IP?
The steps needed to upgrade a network for voice/IP depend on how extensively and where the technology is being deployed.
For local telephony, all external calls still running over the PSTN or over a conventional private voice network using leased lines, so there is no point upgrading existing data communications equipment such as IP routers. The technical requirement is for a new Lan hardware device that routes local calls as IP transmissions over the Lan, and converts external calls back to conventional bit streams.
When considering voice/IP just for local use, ensure that the anticipated savings of running a single set of cables and one local physical network, offset the cost of the new equipment.
Another consideration is reliability, which has to be equal to the conventional private branch exchange (PBX). More commonly, voice/IP is deployed for external calls with the existing on-site telephone infrastructure left intact. In this case, there is a need for gateway hardware to route a call over a wide area IP network. The technical issue here is where and how best to deploy the gateway function. This depends on the size of the site.
One method is to install a dedicated gateway between the existing PBX and the IP network over which the traffic is routed. However, such products are expensive, costing upwards of £10,000, and are more suitable for larger sites.
For smaller companies, there are various cheap options including cards for standard Pentium servers or PCs. One example is Telylink from the Israeli vendor BOS which costs from £935 for a two-line system.
These products must be as reliable as conventional phone switching equipment and also implement the critical H.323 protocol, which is the agreed standard defining the voice/IP encapsulation process and the management of voice traffic within the network.
Another option for larger sites is to incorporate the gateway function within the existing PBX. A number of PBX vendors such as Nortel now offer this. You should consider how well this integrates with the PBX's existing standard telephony functions, and whether features such as call diversion operate seamlessly across the gateway.
Ideally, a PBX-enhanced in this way should be able to route calls over the PSTN or private voice network when the quality of the voice/IP deteriorates temporarily.
Yet another option is to incorporate the gateway function within existing IP routers, which is more appropriate at sites where end-to-end IP telephony bypassing the conventional PBX is being considered.
WHAT'S HOLDING BACK THE QUALITY?
Bandwidth and transit delay affect voice quality during transmission over any communications link. With normal circuit-switched voice, such as the PSTN or cellular GSM radio, the voice is sampled and converted into bits at the rate of 64Kbps. Each bit represents an approximation of the sound during its sampling period, in this case, 1/64000 of a second.
The greater the number of bits-per-second, the more accurately the sound can be represented digitally. Over the PSTN, an end-to-end path is set up, and this imposes a slight delay which is barely noticeable except over satellite links. This delay is very consistent and has no effect on quality. As the bandwidth is fixed, quality does not vary during the conversation. With voice/IP , the digital bits obtained from the sampling process are first packed into IP packets before being transmitted, and this imposes a slight up-front delay. But this is not an issue unless the voice is packed and then unpacked into bit streams during transmissions.
IP networks do not normally provide a fixed end-to-end path for a whole session nor a guaranteed bandwidth. The lack of a fixed path means that IP packets may arrive out of sequence. This is because one packet might take a relatively short path while the next takes a longer route. As a result, packets can arrive in the wrong order, or at least with different time intervals between them, scrambling the voice. This is mitigated by stamping the packets with the time they were sent and using buffers to hold them for a second or so at the receiving end, so that they can be assembled in the right order and with the right timing. But the delay can only be minimal or else the packet will be too late to be of any use to the conversation. Even when there are no serious delays and IP packets have a fixed path, bandwidth can be reduced when there is a surge in traffic leading to degraded quality, unlike with the PSTN IP networks. Instead, everything slows down, meaning that some of the packets have to be discarded to ensure that at least a reduced number can arrive on time. The result is that there are fewer bits to represent the sound at the receiving end and quality degrades. The only way to solve these problems is to set up fixed paths through the network for the duration of a conversation and allocate a fixed amount of bandwidth to it. You create a tunnel through the network shielded from other traffic. This reduces the efficiency of the IP network because this tunnel is then reserved purely for the voice and cannot be re-allocated to other traffic even during periods of silence, until the parties hang up. But it means that a single IP network can carry voice and data, with the resulting cost savings and functional benefits of the technology.
Do you agree?
Have your say on this article