QOS in the relentless pursuit of customers

Only the service provider best able to guarantee service commitments by validating SLA requirements will be able to gain market share

Carriers worldwide today face the same dilemma. While data keeps grabbing bandwidth share of the world's fiber infrastructure, voice traffic still accounts for more than 90% of service provider revenues. Service providers have responded to this issue by integrating all applications on a converged data network - with the result being that "voice" is now viewed as a tricky data application. In this new world order, the most critical roadblock to successful network integration is quality of service.

Service providers, both incumbent and emerging, are under pressure to increase their operating results by gaining market share, expanding services while reducing costs, and creating customer loyalty. Most carriers seem to agree on the strategic directions dictated by the following basic business objectives:

- Increasing billable minutes by optimizing network efficiency

- Reducing customer churn by delivering QOS

- Automating network service management workflow processes by integrating open, best of breed operations support systems (OSSs).

These are general business objectives - the difficulties for carriers lie in the implementation of these objectives. What needs to be clarified are the tactical steps to reach a compelling carrier competitive advantage.

All service providers are looking for ways to differentiate themselves and gain an advantage over the competition. Carriers don't have much room to differentiate on price, so a high level of service becomes the selling point. Difficulties exist because legacy network performance metrics and measurement tools do not support more efficient and automated use of the network while focusing on improving customer loyalty. The traditional legacy network management systems, which focused their data gathering model at internal network operations, are no longer sufficient to achieve these bottom line goals with the required customer satisfaction focus. Such a customer focus is key to survival in today's fiercely competitive marketplace.

With changes in the telephony service and delivery model occurring at record speed, the focus on one constant remains a strong candidate for investment: customer satisfaction. The best tool a carrier has in pursuit of satisfied customers is the service level agreement (SLA).

To differentiate a carrier's SLA, one needs a starting point. New, customer-driven QOS metrics need to be integrated within the QOS management processes for the SLA to be focused on parameters that big bandwidth consumers can relate to. Standardization of these new SLA metrics and OSS integration processes still is being discussed at industry forums such as the ITU-T Quality of Service Development Group and the TeleManagement Forum. With upcoming industry-wide standards, QOS management OSSs will enable service providers to concur with their customers through the provisioning of comprehensive SLAs as well as benchmark themselves. They also will enable carriers to migrate their established voice-based, circuit-switched networks to next generation packet-routed infrastructures.

QOS Management

QOS is a means of describing and gauging network performance. Over the years, network performance, and thus QOS, has been measured in terms of the following four parameters: network availability, error performance, call completion and connection setup time. The dynamics of today's telecommunications industry have created new QOS issues and a stronger need to manage QOS. That includes deregulation, voice and data convergence, Internet effect, multidomain nature of networks, SLAs and the drive to prevent customers from switching providers.

The Telecommunications Act of 1996 had a huge impact on the already deregulated telecom industry. Far-reaching deregulation opened the door to a substantial proliferation of competitive local exchange carriers (CLECs) and paved the way for next generation carriers to operate technology that reduces costs so significantly that QOS becomes the single most critical market differentiation. Only the provider best able to guarantee service commitments, proactively resolve service impairments and satisfy customer expectations with clear and objective measurements validating SLA requirements will gain market share and sustain its competitive advantage.

Conversely, with data traffic overtaking voice traffic, the need for integrated and efficient multiservice networks has given rise to the reality of converged voice, data and Internet application networks. Although the associated advantages of such convergence are apparent - cost reductions, simplification, network consolidation and advanced applications - many issues still need to be addressed. Voice over Internet protocol QOS is seen as the prevalent obstacle to unleash the power of these new technologies. Therefore, QOS will have to address this convergence from the multiservice perspective of voice, video, and data applications and their associated service guarantees, ranging from "best effort" to virtual circuit quality.

It is no secret that the Internet also has had an extraordinary impact on the management of telecommunications networks. The breakup of yesterday's planned and predictable voice traffic in favor of today's erratic, bursty and increasing Internet traffic has led to network planning anxiety and to a host of new QOS challenges. Bandwidth shortages, network capacity management and local loop quality are of critical essence, whereas reliability and data transmission are as important as voice.

QOS challenges are further complicated by service providers' needs to predict how their networks will react under peak loads. This capability is difficult, considering the aforementioned data traffic characteristics and the fact that network usage tends to change dramatically, with Internet access claiming calls that are much longer and distributed all across the day.

Supporting customer QOS management

Lately, SLAs have drawn the attention of service providers who see them as a means of proving to their customers and partners that they can ensure optimal performance of the network services and mission-critical applications, which in most cases now reside on service provider servers. The objectives of SLAs, which can be thought of as network performance guarantees related to QOS, are to define and obligate monitoring of performance levels as well as set and manage customer expectations. By monitoring service levels and providing different levels of guaranteed service, providers can differentiate themselves from their competitors by offering premier services.

One important trend with respect to SLAs is their increasing focus on the customer experience and perception of quality, rather than on the network metrics alone. Service providers know that global network finger pointing and technical jargon debates do not retain customers. Addressing the QOS issue as a customer perception or experience, rather than relying on average traffic performance measures, allows SLAs to be valuable to the service provider and the customer. Service providers get the benefit of being proactive with customer expectations. Meanwhile, customers get a guaranteed level of service benchmarked against parameters they can easily relate to their day-to-day experiences.

The TM Forum has recognized that trend in its focus on managing customer expectations and its proposed framework for customer QOS management that binds network operations with customer care processes, which encompass monitoring, managing and reporting of QOS as defined in provider/customer service descriptions, SLAs and other service-related documents (Figure 1).

While the TM Forum focuses on how to manage customer QOS and SLAs using an OSS framework, the ITU-T's QOS Development Group focuses on developing new metrics, therefore laying the foundation for objective, measurable SLAs that reflect QOS from the customer's viewpoint. New recommendations such as P.861 (Perceptual Speech Quality) and E.425 (Network Effectiveness Ratio) are good examples of the push to improve.

The drive for customer QOS-based SLA management dictates that the performance management capabilities of integrated voice and data networks must evolve. While the impact of the Internet is such that voice is now considered as a "tricky" data application, many questions arise to ensure that voice still is transmitted with, at the least, near toll quality. In applications where the transmitted voice is half-duplex such as streaming audio applications, network impairments and degradations can be handled using packet buffering, retransmission protocols and correction algorithms.

This strategy cannot be employed, however, when the application is full-duplex real-time voice, such as IP telephony applications in which the small delay budget must be met and small jitter buffers do not allow for packet retransmissions. Moreover, people are not computers and typically do not like to retransmit information over the phone. Other voice conversation characteristics that will need greater packet network service assurances relate to packet loss compensation, echo cancellation, call clarity, network availability and call completion. Furthermore, the successful implementation of QOS for packet-routed voice will open the door to more advanced applications such as mass videoconferencing, which needs the same types of quality assurances as voice, but with much greater bandwidth capacity.

QOS metrics are improving

What is driving service providers to improve QOS metric gathering and usage? It is the trend away from internal operations to a new customer perception focus. The quality monitoring model has changed from operations and network departments toward the customer and marketing. It has also changed from a hardware measurement to a software OSS one. These forces are driving network service and optimization changes back to the carrier now more than ever.

Judging from the debates during the ITU/QOS Development Group meeting held in Budapest last September, QOS metrics must evolve to encompass new traffic types, convergent networks and changing telecom business models. While the answer seizure ratio (ASR) used to be the only measure by which carriers would rate themselves and exchange traffic, the ITU/QOS Development Group now has fully recognized the customer-focused SLA trend and has taken evolutionary steps to define new parameters and methods that characterize network performance.

The direction is to have a mixture of QOS metrics that are driven by customer perception and the bottom line. First, new metrics need to be defined to account for the customer's perception of QOS. Then, bottom line metrics such as ASR need to be redefined to focus entirely on the service provider's own network segment.

In the bottom line-driven category, the ASR definition has been updated to strictly reflect the network's ability to carry billable minutes in a multicarrier environment. This was achieved by filtering all user related events such as ring/no answer, misdialed numbers, line busy and so forth from the ASR metric, thus generating a new metric called network effectiveness ratio (NER).

Such a measure is extremely important to service providers because it has a direct effect on their bottom line and provides a metric that is extremely focused on their own network segments, regardless of the overall QOS effects of other networks involved in the same digital path. For example, ASR results traditionally tend to top off at around 70%, while NER results, taking caller and receiver behavior into account, have greater accuracy for determining true network-related performance gaps and can top around 95%.

Within only a few months, the ITU normalized a set of user-focused QOS metrics called "average length of calls" (ALOC) and "post gateway answer delay" (PGAD). These metrics are call data record-based and enable service providers to easily benchmark traffic routes together in terms of the customer experience - without having to deploy costly test and monitoring equipment. Average length of calls is based on the fact that a high-quality route enables customers to relax on the phone and thus hold longer conversations, whereas a poor quality route will have the opposite effect. Post gateway answer delay provides a way to keep track of post dial delay using ubiquitous CDR data. Post dial delay is a very annoying factor for customers and a direct revenue loss for service providers.

The other development path for improved QOS metrics is perceptual speech quality measuring. There are two approaches to determining perceived voice quality: parameter-based and speech-based. In the parameter-based approach, the objective is to measure relevant network parameters, correlate them with the associated customer impairment perceptions using a mathematical model and obtain a prediction of subjective mean opinion score. The objective of the speech-based approach is to inject speech samples into the network under test, analyze the received and degraded speech waveform against the known reference speech using a mathematical model and obtain a prediction of subjective mean opinion score. Table 1 provides a comparison of both methodologies.

These types of measurements will become more important as voice traffic transmissions span several transport networks in tomorrow's competitive, least-cost routing industry. The perspective of voice-over-IP traffic spanning several carrier networks contrasts heavily with the tariff-regulated markets that still exist today. Not long ago, one could assume that a call from the U.S. to the U.K. was routed through AT&T and BT's networks. The very near future can have many scenarios - some involving traffic routing through up to a dozen or more service providers. That is why service providers need metrics that provide visibility of their own network and the end-to-end user-perceived measurements.

The need for QOS standardization

With the proliferation of competitive service providers all keen on implementing QOS as a means for differentiated services, service performance and customer satisfaction could gain from the adoption of a common industry language such as a quality index.

Such quality indexes already exist, such as the call quality index, which is a combination of five QOS metrics. These factors include signal level, noise, echo path loss, echo path delay and speech activity. The call quality index is expressed on a scale of 0 to 100 based on a technical model provided in ITU-T Recommendation G.107. Call quality index scores above 90 are considered "best" while call quality index scores below 60 are considered "degraded."

The limitation of simply reporting index scores, such as mean opinion score or call quality index, is that they do not provide carriers with the information required to pinpoint and recover from problems before customers start complaining or switch carriers. Fortunately, OSS technology is available today providing mean opinion scores, customer satisfaction levels and relationship to impairment sources (Figure 2).

This energy spent on bringing SLAs and QOS to multi-traffic multiprotocol networks will serve to better integrate these networks with wireless networks. However, this is little reassurance when considering all the possible applications of future wireless devices and services. Already, much talk is about tomorrow's network - a wireless high-speed data access network linked to an optical fiber backbone running the IP suite of protocols. The anticipated growth in consumer devices and Internet services also stresses the need for efficient network performance perception and measurement to better manage service quality.

Service providers that don't take the challenge will miss out on the market's extraordinary revenue-generating business opportunities simply because of a lack of market acceptance due to unpredictable service levels. The challenge thus becomes clear: Customer QOS management through the development of a common SLA-based environment is the competitive weapon for gaining the highest-bandwidth customers for voice, fax and data services while retaining current customers and optimizing constantly changing network investments.