 | Issue: | Latin America I 2001 |
| Article no.: | 7 | |
| Topic: | IP Telephony Quality-of-Service Aspects IP Protocol – Changing the Paradigm | |
| Author: | Oscar Avellaneda and Bruce Pettitt | |
| Title: | Not available | |
| Organisation: | Nortel Networks | |
| PDF size: | 36KB |
About author
Not available
Article abstract
The first IP Telephony applications were inexpensive, low quality alternatives to traditional service. Convergence towards integrated multimedia, multi-service networks is driving the move to a packet-based telecommunications infrastructure. Integration of voice and data onto a single network infrastructure offers significantly improved efficiency for both private and public network operators. The future of IP Telephony depends, in good part, upon the quality of service that can be provided using this technology.
Full Article
The deployment of networks using the Internet Protocol (IP) is becoming increasingly significant in the world of telecommunications. The transmission of voice, fax and related services, either wholly or partly over IP-based packet-switched networks, is a key consideration in planning new telecommunication networks and in expanding existing networks. This article addresses Quality-of-Service (QoS) aspects which need to be considered in implementing such IP Telephony solutions. When basic forms of IP Telephony such as Voice over IP (VoIP) were first introduced commercially in the mid-nineties they were marketed as inexpensive alternatives to traditional telephone service. QoS was not a critical issue; users were willing to tolerate low quality because of the low cost. Today, the prices of traditional telephone services have dropped dramatically and the low-cost of these basic forms of IP telephony is no longer as appealing. The technical mechanisms for carrying voice-over packet networks have been evolving quickly in the last several years as have the business models adopted by network operators and service providers around the world. Convergence towards integrated multimedia, multi-service networks is driving the move to a packet-based telecommunications infrastructure. Integration of voice and data onto a single network infrastructure offers significantly improved efficiency for both private and public network operators. Together with the fundamental technology capabilities, emerging standards are key to the successful implementation of the next-generation converged network, which will integrate existing voice networks with data networks. Indeed, the proliferation of the suite of IP standards supporting a general-purpose communications infrastructure, combined with the significant developments of recent years in networked multimedia, including IP Telephony, have created the need for serious consideration of inter-operability of newer IP-based implementations with the legacy Public Switched Telephone Network (PSTN). The fundamental IP standards have been developed by the Internet Engineering Task Force (IETF), and related standards are being developed by other standards bodies — including the ITU’s Telecommunication Standardisation Sector (ITU-T) and the European Telecommunications Standards Institute (ETSI). First of all, from the Quality-of-Service perspective, it is worth noting that the original “best-effort” internet was designed to guarantee network survival in time of physical disruptions — with a view to ensuring the retention of maximum connectivity following severe network failures. The Internet was not designed to support high-speed, real-time applications in a relatively stable environment. In today’s networks, with highly reliable network transmission facilities, new mechanisms can be deployed in IP-based networks to support real-time, mission-critical, interactive applications. Quality-of-Service/Performance guarantees can also, now, be described in quantitative terms in Service Level Agreements. Certain challenges remain in fully integrating newer IP Telephony implementations with the traditional telephone network and in ensuring harmonious inter-working. From a QoS perspective, technical solutions are available to ensure meeting end-user satisfaction requirements and, at the same time, ensuring more efficient, cost-effective, operations for network and service providers. Nonetheless, careful assessment of the detailed service requirements and network environment, in a particular situation, remains quite complex. End-user Perspective Addressing QoS requires a good understanding of the needs of end-users of the relevant services, as well as an end-to-end understanding of the network and the various operational factors which can affect service quality. What are some of the factors which impact voice service quality as perceived by end users? End-to-end call-setup delay, voice-transport delay, and delay variation (or jitter) in the delivery of data are among the fundamental performance considerations. Other parameters relate to loss of data, either because of transmission errors (which are very rare in today’s digital networks) or through packets being lost because of congestion. In terms of delay: as one looks from left to right, the applications are increasingly tolerant of delay impairments. For example, interactive games are highly delay sensitive, and even web browsing and e-mail activity is sufficiently interactive that delays of the order of a few seconds are the maximum which users feel are tolerable. By contrast, while faster is always better, users are more tolerant of delay for the transfer of files, images, and for paging or bulletin-board type applications. With the merging of networks (shown in Figure 2), there are a number of components of end-to-end delay that need to taken into account. The total delay (and delay jitter) from all these components needs to be smaller than specified performance objectives. The different stages of the voice path will contribute to the total delay (coding, packetisation, queueing, and build-out). For the present discussion, the QoS challenge can be posed as that of providing support for IP Telephony in a performance envelope of reasonably low delay and minimal packet loss. Roughly speaking the constraints are end-to-end delay not more than 100 to 200 milliseconds and packet loss not exceeding about one percent. The overall challenge for operators and service providers is to manage the network and access channels so as to provide highly reliable, real-time interactive telephony service while still providing adequate performance for other applications. In supporting the evolution of IP Telephony, a great deal of effort has gone into assessing the performance of various speech-encoding techniques and the responsiveness of the network in initiating and setting up calls. It has also been necessary to develop supporting signalling protocols to facilitate call setup and appropriate resource management to ensure meeting performance objectives and satisfying end users. It is important to discuss some of the key technology enablers of QoS for IP Telephony. First, it is important to note the advances achieved in the speech-coding techniques which make possible more efficient use of bandwidth while maintaining high-quality speech transmission. Complementing these fundamental speech-processing capabilities are a number of network-oriented technologies — notably generic Traffic Engineering, Transmission Planning, QoS-Resource Management methods and related QoS Routing, Signalling, and Control Mechanisms. Speech-coding Techniques For transmission over today’s digital networks, an analogue voice signal is digitised using a codec — a coder-decoder. Codecs convert the analogue voice signal to digital and, at the far end, back again. Table 1 shows some of the characteristics of the codecs most commonly chosen for point-to-point voice-over-IP applications. Traditional digital telephony is based on a technique known as companded Pulse Code Modulation (PCM) which provides analogue-to-digital conversion for the audio channel. Details are found in ITU-T Recommendation G.711. Through the use of more complex algorithms, encodings at lower bit rates are feasible — and reasonably good performance can still be achieved. Examples of lower-rate codecs used in IP telephony applications are defined in ITU-T Recommendations: G.729, G.726 and G.723.1. Note also the GSM codec developed by ETSI. Although lower-rate encoders utilize less bandwidth, their use does entail additional delay — because a sample, a frame, of a given length must first be accumulated before it can be processed by the encoder to produce a compressed version. At the receiving end, there is further delay in receiving and decoding the frame into speech before it can be heard. Listening tests have shown that the perceived quality can remain quite acceptable even when using a much lower rate codec (for example, the G.723.1 codec at only 6.3 kbit/s). Wideband can also be encoded using techniques that enable a broader range of frequencies of the audio source to be encoded. This produces what is sometimes referred to as Broadcast-Quality speech. High-fidelity, wideband codecs using various algorithms operating at higher bit rates (such as 128 kbit/s) can provide excellent ‘broadcast quality’ audio. Today, newer codecs (for example, the G.722 codecs) provide wideband quality at 64 kbit/s and below. Network Aspects Speech encoding/decoding takes place either at the end-user’s telephone, his computer, or at an access gateway. Let us turn now to some of the more network-oriented aspects of QoS. Echo Control and Transmission Planning Echo, which impacts voice telephony in a communication network, results from coupling, a crossover, between the transmit path and the receive path which causes the outgoing speech to be sent back to the talker. The audibility of an echo depends on two factors: the amplitude of the echoed signal and the time it takes to return to the talker. Consequently, the delays which can arise in packet-based networks (and the delay variability) means that special care must be taken with potential echo impairments. Solutions for echos typically involve the use of echo cancellers or echo suppressors. Traffic Engineering In a narrow, more traditional sense, traffic engineering is an aspect of network planning which involves network dimensioning and capacity management to meet forecasted loads over the long-term. In IP-based networks, the term “Traffic Engineering” can best be described as the performance optimisation of operational networks. It connotes not only the measurement, modelling, characterization and control of traffic loads over time-scales in the order of days or weeks, but also addresses call and connection routing, network resource management, the management of routing tables and dynamic transport routing techniques requiring the handling of traffic loads over much shorter timescales. QoS Resource Management In order to manage network performance so as to meet QoS requirements, various techniques have been developed — and are still being refined. These include the introduction of differentiated service categories and the use of priority mechanisms (such as handling a packet at higher or lower priority, depending on its QoS class). This can involve multiple priority queues and/or techniques for discarding certain lower-priority packets under congestion. Various mechanisms to reserve resources have also been developed. These ensure the availability of the bandwidth needed to meet a service request. The virtual networks, often called a virtual private network or VPN, are another way of providing assured capacity for a particular user or group of users. These techniques, together with supporting signalling and flow-control protocols, manage network resources to ensure that the end-users of IP Telephony’s voice, fax and related services will enjoy appropriate Quality-of-Service levels. The Outlook for the Future Today, many tools certainly exist for addressing specific aspects of QoS for IP Telephony, but the solution in a particular case can depend critically on: – end-user service specifications; – the need for inter-working with other networks; – possible requirements for integration with legacy infrastructure; and – the size and complexity of the network. Moreover, the emerging QoS standards for IP Telephony need further refinement, and it must be recognized that, in general, network design remains quite complex. It must also be noted that opinions vary on the extent of this complexity — with some people arguing that the simple expedient of adding bandwidth (at increasingly low cost) solves many of the QoS issues — while others argue that there is a need for more sophisticated, fine-grained, call-control methods. Furthermore, access through wireless, cable, or digital subscriber loop systems can present various particular special requirements. To summarize: for the successful, widespread deployment of IP Telephony, it is essential to develop the right mix of tools for ensuring that end-user service requirements will be satisfied. Conclusion Exactly how this is best done from either a purely technical or a cost-effective perspective will likely be a subject of debate for some time to come. In the near term, debate will continue over the relative merits of ‘over-dimensioning’ versus ‘traffic-engineering/QoS’ solutions. And in the longer term, a more fundamental question to be addressed is whether emerging QoS solutions can fully enable multimedia convergence with an IP-based infrastructure.
