Fierce VoIP
July 22, 2005

Quality of Service: The Role of VoIP Silicon and Software

By Majid Foodeei

Over the past two years almost all of the regions in the world have witnessed massive VoIP growth. The threat of low-cost VoIP offerings from new competitive players like Yahoo, BB in Japan and Vonage in the US have served as a catalyst for consumer VoIP implementation. This trend has been followed by incumbent global telecom service providers in search of new revenue unveiling large-scale plans of their own. Most recently, the small but disruptive peer-to-peer VoIP player Skype announced a rapid subscriber adoption of about 30 million and counting.

Classical VoIP and more general voice-over-packet markets encompass many existing and emerging applications like enterprise and wireless. Boundaries between technologies like cellular and WiX are becoming blurred by network convergence and next-generation network evolution. The delivery of solid end-to-end voice quality or QoS across the evolving heterogeneous networks is now met with renewed challenges.

Today's QoS Challenges
From access to core, VoIP has many faces. It is challenging to deliver end-to-end VoIP quality over a patchwork of networks with both old and new equipment. This classic VoIP obstacle was widely believed to have been resolved, but it has recently reemerged. As an end measure, voice quality and QoS become especially challenging as they traverse heterogeneous networks, dissimilar codecs and piece-wise design objectives that comprise the large underlying system that a VoIP call uses.

Classic QoS implies the enabling of preferred service quality to selected traffic (in this case, voice). In order to ensure end-to-end quality of voice/service, which is built on multiple "clouds," network components and infrastructures, the QoS features must be available and work together across these network components. Thanks to the accuracy of human auditory perception, voice quality is extremely time- and packet loss-sensitive. This sensitivity does not exist for other traffic (such as data) concurrently exists with voice. Current QoS methods used include the following:

· MPLS or preferred classification applied in layer 2/3: Class of service in layer 2 and differentiated service DSCP in layer 3.

· Traffic shaping: Smooth transmission to minimize packet loss by limiting I/F traffic to average/peak limits and policing, meaning the handling of voice traffic in a preferred manner.

· Classic queuing for congestion control: Whenever there is a burst of data traffic competing with voice in a "pipe," voice gets priority; techniques such as weighted (priority) fair queuing are used.

These techniques are usually implemented at the network's edge, aggregation points, size mismatched pipe junctures and flows where voice and other traffic are mixed. The following practical VoIP network planning methods have helped to deliver QoS for some time:

· Single vendor or rigorously tested interoperating multi-vendor network. subsystems (exact choice of equipment and software releases).

· Overengineering: Extra bandwidth and over-capable equipment (cost).

· TDM and guaranteed bandwidth connection-oriented usage (e.g. ATM VCC).

As network convergence and global usage of VoIP become a reality, ensuring quality using mere classic QoS techniques is simply not possible. This is mainly because voice calls traverse non-interoperable equipments across the globe and pass through multiple clouds (wireline, wireless telephony and data networks, edge, and core). Multi-vendor equipment cannot be rigorously pre-tested on an ongoing basis. National boundaries introduce various network and quality threshold conditions. In addition, packet network advances and architectural revisions, enabling many new services, reducing cost (commoditization) and increasing capacity have brought new QoS problems and challenges. To meet these challenges, extensions and new techniques beyond classical schemes have been invented to emphasize the robustness of architecture evolution and adaptability. More than ever these techniques need to be available across all traversed network components to enable end-to-end QoS.

Silicon and Software in QoS
Advanced VoP silicon and integrated software ("gateway-on-a-chip" or VoP SoC) have been available for some time. The VoP SoC's first requirement is to continue furnishing voice quality features provided by previous-generation DSP despite fundamental changes in the converged network environment (at lower cost). Moreover, interoperability and robustness across heterogeneous next-generation and wireless networks are expected. The following is a non-exhaustive list of VoP SoC features that play a major role to enable VoP voice quality and QoS:

· Enabling QoS primitives such as ToS bits, DiffServe and MPLS at VoP silicon level

· Basic voice quality and seamless voice-, fax-, text- and modem-over-IP/ATM/TDM

· Carrier-class echo cancellation .

· Lowest possible delay (one example: VoP SoC end-to-end delay including jitter buffer is required to be no more than 25 ms).

· Simultaneous support of wireline and wireless codecs and protocols.

· Very low-jitter VoIP packet traffic generation by VoP SoC (this will reduce end-to-end jitter-related voice quality degradation).

· Interoperability with a majority of implementations (this will reduce interoperability issues at equipment, network planning and deployment).

· Adaptive jitter handler, PLC and other quality techniques (these features and others are to perform under unpredictable heterogeneous network conditions).

· Voice enhancement features such as noise suppression, adaptive gain control and acoustic echo control.

· End-to-end transcoding and tandem/transcoder-free operation (TFO/TrFO).

· Voice quality monitoring (e.g., RTCP and RTCP-XR) and diagnostics. This is key in identifying trouble spots in heterogeneous converged networks.

Meeting the Challenge
Seamless service across heterogeneous ATM/IP/TDM wireless and wireline networks requires network transcoding and interworking with full network awareness, paying special attention to equipment interoperability. Today's advanced VoIP silicon and software meets this challenge by offering chip-level network interworking with and without transcoding as well as interoperability.

The evolving NGN VoIP network includes an unprecedented variety of wireless, wireline and next-generation terminals. The impact on voice quality due to acoustic characteristic, volume variations and background noise pick-up raises the need for more robust voice enhancement features on silicon. Advanced silicon software in particular is now expected to have robust noise suppression, adaptive level control and acoustic echo cancellation/suppression that can handle a variety of terminals and networks.

Vendors, providers and end-users can all benefit from the current boom of VoIP. However, this is on the condition that they meet the challenges of end-to-end voice quality across heterogeneous networks. Some key technology enablers include advances in classic network QoS that are implemented across all network components, as well as quality and QoS features available by advanced VoIP SoCs.

Without advanced VoP silicon and software-enabling technology, ensuring voice quality and QoS is a costly, complex and slow process. NGN VoIP silicon and software designed from the ground-up to meet the challenges addressed in this article are available today.

Majid Foodeei, is responsible for system architecture and technical marketing for Centillium Communications' Entropia product family.