Bridging the VoIP Disconnect Between Users and Operators

There’s an amazing opportunity on the horizon for VoIP service providers to distance themselves from their competition if they pay attention to the measure of service quality that matters most: the subscriber perception of quality. Opinion polls and surveys can be used to garner this data, but a much more comprehensive solution focuses on actually measuring each and every call and scoring calls individually using sophisticated quality-modeling algorithms. Cumulative reports based on this data can be used to analyze performance for an individual subscriber or an entire subscriber population and service offering. Embedding these measurement tools in the endpoints of the IP network makes comprehensive measurement and reporting possible.

But IP networks introduce a disconnect between operators and their subscribers. That makes it difficult for the operator to understand and measure the subscriber’s perceived service quality. This disconnect stems from the fact that measuring, monitoring and analyzing the quality of service delivered via VoIP requires a different approach and a new mindset from the way the established circuit-switched PSTN has been managed in the past. Comprehensive network management measures for the PSTN have been in place for decades. Operators are able to monitor and record exhaustive data on every aspect of the network’s operation right down to the number that was dialed and when a line goes off-hook. Dealing with an IP-based packet network is not so easy.

There’s good reason why virtually every graphic illustration of an IP network places a cloud between the service provider and its end users. Whereas straight lines could depict a call over the circuit-switched network quite accurately, there’s no way of knowing or even predicting the path or paths that a packet-based call will take over an IP network. Cloudy is the best way to describe the path that an IP call takes.

Other complicating factors are the distributed nature of IP networks and the migration of intelligent functionality to edge devices. Embedding measurement in these edge devices restores the visibility that the IP cloud obscures. The structure of the PSTN is straightforward and hierarchical. Analog telephones are very simple devices. The intelligence in the PSTN is concentrated in central offices and cascaded to a lesser degree to intermediate nodes.

In contrast, an IP network has a decidedly horizontal structure. Functionality has been moved outward away from the purview of the CO to customer premises equipment like residential gateways, IP phones and phone systems, home networks and various broadband access devices. At the same time, the IP network infrastructure itself has been populated with new types of systems such as session border controllers, proxy servers and media gateways. Any of the intermediate steps along the path of an IP call can impact the quality of that call. And the presence of these intermediary steps can become a barrier to the visibility that operators need to manage the quality of their subscriber experience.

To better manage the quality of the service provided by their networks, IP network operators have struck upon several methods that provide a modicum of visibility into the operation of their networks. Unfortunately, these methods are partial solutions at best. They offer only a snapshot of the network’s performance – not the total visibility to which telephony operators are accustomed.

  Enlarge

Some service providers have installed hardware probes on a portion of their IP links as a way of gathering information on the performance of services utilizing these connections. This is an appropriate way to measure the performance at centralized concentration points such as at a session border controller or SIP proxy. Unfortunately, the cost of these probes makes it prohibitively expensive to install this hardware on every link in the network and certainly prohibitive to install probes at the subscriber residence. As a result, information gathered from only some of the links in the network forms the basis for an extrapolation on the performance of the entire network. From this, a further extrapolation leads to a measure of the subscriber experience on the network.

Another way to gauge the quality of the subscriber experience involves test calls placed by the service provider. These test calls can be monitored and the results used in an extrapolation of the quality of service provided by the network. They have the advantage of being directly measurable because the injected signal (voice file or tone sweeps) is known and direct comparison of the input and measured signals can be made. Of course, test calls do not represent perfectly the subscriber experience because they are placed from a controlled environment rather than the subscriber location. And the extrapolation injects a certain degree of approximation into the process as well.

To overcome the shortcomings of these techniques and to complement their benefits, new tools are being embedded into IP technologies at the component level. This new generation of quality measurements will gather real-time and complete information on the operations of the network as well as the quality of the subscriber experience provided.

 Enlarge

Embedding quality management elements into the technology deployed at the endpoints in IP networks avoids costly hardware probes because these measures are integrated at the component level and can be deployed readily by equipment manufacturers. These embedded measures give service providers pervasive monitoring coverage where they need it to identify user-quality problems such as dropped calls, calls-not-completed, voice-quality impairments due to packet loss or jitter, echo and other shortcomings.

This type of comprehensive information could be provided to a Mean Opinion Score (MOS) algorithm via a standard interface such as the Real Time Control Protocol with Extended Reports (RTCPXR). A MOS algorithm could score the quality of each call on the network and compare this information with predefined standards for acceptable service quality. Any deviation from the quality threshold could be stored in a database and used as the basis for targeted maintenance and repair activities. Or, a deviation from the acceptable quality level could trigger alerts to technicians for further investigation. Some sophisticated service providers might use this information to deploy methods by which the network itself could perform self-healing actions to mitigate service impairments.

One such self-healing action might involve automatically changing a piece of equipment’s operating parameters, such as the coder/decoder (codec) implemented in an IP phone system or residential gateway. The typical PSTN-quality voice codec deployed in VoIP equipment these days is G.711, a fairly bandwidth-intense codec. It specifies a packet payload of 64kbps, but overhead brings the total bandwidth requirement to around 80 or 90kbps.

If embedded quality management in a CPE system detects that packets are being lost, the system can switch automatically to a more robust mode of operation by implementing an alternative codec like G.729. With a packet payload size of 8kbps, G.729 is less bandwidth-efficient than G.711, but a lost G.729 packet will have less effect on the quality of the subscriber experience since less data is lost. Moreover, because G.729 packets are smaller, packets can be duplicated and transmitted two or three times. Then, if a packet is lost, chances are still good that a duplicate will be received and the quality of the subscriber experience will be maintained at a high level.

The term “quality of service” originally referred to the operational need for packet-based networks to guarantee or dedicate bandwidth for time-sensitive applications such as voice conversations. But if the concept of QoS is expanded to include every aspect of the quality of a subscriber’s experience with a VoIP service, then QoS takes on even greater importance for operators. By concentrating on this broader construct of QoS, the IP service operator maintains the proper focus on the quality of its subscriber experience.

Tom Flanagan is director technical strategy of DSP Systems for Texas Instruments Inc. He can be reached at tflanagan@ti.com.

Texas Instruments Inc. www.ti.com

Share this article: Email, Slashdot, Digg, Del.icio.us, Yahoo!MyWeb, Windows Live Favorites, Furl
Add this article feed to: RSS, My Yahoo, Newsgator, Bloglines