Migrating to a software telephony architecture

Not too long ago, a TDM-based service node connected to Class 5 circuit-switching platform could cost millions of dollars. Now, utilizing open architecture software and networks, the same capabilities can be built on a fully-redundant Linux hardware footprint that costs less than $25K. Instead of costly circuit-switched, single-application platforms, carriers can deliver multiple services on a single applications server or cluster of distributed servers, utilizing Session Initiation Protocol (SIP), which has emerged as the industry standard for Voice over IP applications and other media transmitted as packetized data.

Application servers based on a pure-IP design with integrated SIP and Media Gateway Control Protocol (MGCP) support allow developers to design applications that interact with media gateways, IP phones, SIP user agents and IP media servers. This enables an entirely new approach to deploying enhanced services in converged TDM/IP and VoIP networks, where services can be added and reconfigured literally on-the-fly utilizing the same equipment with no down time.

This revolution in communications applications relies on open architecture and use of next-generation network standards—including SIP, VoiceXML, MGCP, and XML—to provide a multi-service platform with tremendous flexibility and scalability. This “next-generation network” enables distribution of call processing across many application servers that can be configured in a high-availability cluster, co-located or distributed across an IP network. Load balancing allows an integrated SIP proxy server to direct calls to the least busy Application Server in a cluster, resulting in a single service access point comprised of a cluster of application servers able to scale up to well over 50,000 simultaneous calls—enough to process 1 billion minutes of prepaid calling card minutes a month in a single node!

Historically, the market for carrier-class enhanced services has been a struggle to innovate. As measured against the demand for services, market growth rates have been moderate, the number of vendors providing solutions has been small, and the cost of creating and delivering solutions has been high. Most vendors provided closed, proprietary and non-interoperable solutions. This fragmented the market and reduced the choices available to service providers. It has also presented service providers with a range of problems, including long development cycles; high maintenance costs of supporting multiple, incompatible platforms; and poor integration of services across platforms.

The traditional approach used to deliver high-value voice services to PSTN networks was through the deployment of service node platforms which included dedicated, purpose-built hardware and software capable of deploying services such as voice mail, find me/follow me, unified messaging, prepaid calling card, interactive voice response, 800 number services, and conference calling. Most of these service node product offerings were designed to deliver only one of these services at a time using a proprietary design that made it difficult to deploy multiple services. If service providers had a requirement for new services, service node vendors typically developed customer-specific software releases that could be very costly. This software development process typically took six to 18 months since service logic was tightly integrated with internal signaling subsystems, proprietary DSP hardware and capacities of underlying hardware components.

A completely new approach has emerged to deploy a next-generation services architecture using IP-based technologies. This approach decomposes the traditional TDM service node into hardware and software components separated into distinct functional subsystems and interfaces based on open communications protocols—SIP, MGCP, Real-time Transport Protocol (RTP) and commercially available computer systems, databases, and web servers. Service logic is now separated from the hardware and runs on what is generically called an application server. Newly-developed XML-based Service Creation Environments (SCE) generate service logic that is separated from underlying hardware platform dependencies, providing a clear delineation of hardware and software components allowing service providers to specify best-of-breed components. 

Another major component of this next-generation services architecture is the IP-based media server, which takes DSP logic that ran deep inside traditional service nodes and makes these functions available to the service logic execution environment running on the application server. These functions include interactive voice response capabilities such as playing prompts and collecting DTMF digits, automatic speech recognition (ASR), bridging calls together for conferencing applications, as well as recording greetings, announcements and messages to a server. As call flow logic executes, the application server requests these functions when needed from the media server using either SIP or MGCP protocols. This decomposed service node model speeds up application development and allows components to scale independently so that new services and capacity can be easily added.

Fitting these components into existing or greenfield networks is a relatively simple matter of adding a media gateway or voice router that supports SIP signaling, then connecting the application servers and media servers behind it. 

Emerging carriers can deploy media gateways to perform IP trunking that provides connectivity to the PSTN and allows for transport of voice traffic across IP networks. This creates a very cost-effective approach to build out brand new networks utilizing IP network infrastructure, but it also enables them to incrementally add IP voice services to this network by “plugging” in SIP-based application servers and IP media servers to the converged network infrastructure. This provides significant cost savings versus implementing TDM service nodes, speeds up time-to-market and provides services definition flexibility. This is especially important to emerging carriers who are competing with incumbent carriers on the basis of lower pricing and service offering differentiation.

Software-based media servers can be deployed as a stand-alone IP-based media server solution to support SIP entities throughout a SIP-enabled network or in combination with an application server running SIP-based services. Software-based media servers can support a variety of media processing functions, including announcement generation, DTMF detection and generation, message play and record, conference recording, audio bridging  for small n-way conferences, and other advanced capabilities. These functions can be logically combined and embedded in a service logic execution environment to implement a wide variety of basic and enhanced services, including playing announcements, calling card, conference calling, interactive voice response, and voice messaging, as well as custom-built applications. We’ve been able to support to 400 full duplex IVR sessions with DTMF detection on a single processor using the software-based media server approach.

Because it is possible to run these software-based media servers on industry-standard hardware running Linux, service providers are able to benefit from a low cost-per-port. As faster hardware becomes available, carriers can scale capacity without expanding their hardware footprint and without disrupting services. Common hardware components also means hardware spares can be reduced, eliminating the need to inventory dedicated media server platforms. Carriers have the added flexibility of choosing a variety of cost-effective Linux platforms.

For traditional carriers with a large investment in TDM networks and service nodes, IP voice services can be deployed today to replace TDM service nodes, creating a bridge between the PSTN and an IP network that enables these systems to be accessed by service subscribers on any wireline or wireless phone. Since this offloads voice traffic that is specific to services onto an IP-based telephony network, this deployment scenario is called “enhanced services offload”. This approach allows service providers to either replace legacy TDM-based services or cap-and-grow onto a very cost-effective services architecture.

Let’s look at a couple of examples: In the case of prepaid calling card, a subscriber would typically call an 800 number provisioned on the media gateway to get access to the service. Based upon the number called, the media gateway “forwards” the call to the application server by issuing a SIP invite request. The application server then starts prepaid service logic that plays an introductory service greeting and prompts the subscriber to enter their PIN code. During this part of the call flow, the application server asks the media server (usually under MGCP control) to play the prompts—“Please enter your PIN code”—and collects the subscriber’s PIN code that was entered using DTMF input on the subscriber’s phone.  The “bearer channel” or voice path that enables the subscriber to hear the voice prompts is established between the media gateway and media server as RTP streams. 

Conference calling has traditionally been very expensive to deploy using TDM approaches. Typical TDM conference bridges today might consume an entire 19” rack with maybe a 1,000 ports, whereas IP media servers can physically scale up to 18,000 ports in a single shelf; this greatly reduces carrier costs for network connectivity, maintenance, and manpower. Given the IP nature of these new services platforms, Internet integration for subscriber control of the conferencing service is much simpler to integrate architecturally. For conference calling, subscribers call the 800 number on the media gateway to access the conferencing service.  Again, subscribers are authenticated as in the prepaid calling card service example. The application server creates an active call session for each participant in the conference call and maintains the state of the conference. All of the audio from these call sessions is bridged on the media server, and RTP streams are maintained for each participant for the duration of the call. The application server can be processing both of these services at the same time, allowing the service provider to consolidate resources and manage multiple services in a single services environment.

While the Internet explosion derived from the standardization of a single subscriber interface (the HTML-based browser) and a single communication protocol (HTTP), standardization at this level is not possible in the telecommunications arena. Telecommunications solutions must allow subscribers to communicate over an ever-expanding variety of physical and network interfaces (landline, cell phone, pager, browser, handheld, etc.). For the foreseeable future, enhanced services platforms will need to "speak" a variety protocols and support a burgeoning portfolio of services.

XML is an open, extensible service description language and execution framework well-suited to providing a standard way of describing and delivering next-generation, network-based enhanced services because it is inherently extensible and scalable. Open extensibility allow vendors or groups to extend the language to describe the capabilities of new technologies, protocols or interfaces in agreed-upon ways. Using XML, we were able to create a service description language and associated service execution framework that allowed us to provide sophisticated, third-party call control capabilities to SIP-based networks. The result is a Service Creation Environment and application server that provides a multi-service platform for creating both line-side and/or trunk-side network-based services with the speed and flexibility demonstrated in delivering XML-based Internet applications.

Using this XML-based approach, developers are able to quickly and easily build applications using drag-and-drop components to enable various built-in operations, linking them in a visual flow-chart style representation to construct the call flow. A comprehensive set of built-in event handlers can be specified as call flow functions to trigger logic at any point during a call. Developers can add C, C++, or Java code to incorporate their own programming logic into a call flow. Two other notable efforts have been made to define XML vocabularies for telephony-related application domains: VoiceXML, a vocabulary for describing speech-enabled interactive voice response dialogs, and Call Control XML (CCXML) to control the setup, monitoring, and tear down of phone calls.

While the standardization effort is still underway, it is clear that we have made tremendous progress in migrating from a hardware-based to a software-based architecture. Manufacturers of circuit switches are rushing to build or acquire the expertise to bridge this transition; TDM-based service providers are capping investment in expensive hardware-based architectures and learning how to integrate the new software-based architectures to leverage existing infrastructure while gaining the ability to quickly and cost-effectively roll out next-generation IP enhanced service solutions for converged TDM/IP networks. The open architecture, software-driven approach to telephony is pushing down the cost of delivering enhanced subscriber services while enhancing carriers’ flexibility to innovate and adapt.

Ken Osowski is vice president of marketing and product management of Pactolus Communications Software Corp. He can be reached at kosowski@pactolus.com.

Visit Pactolus Communications online.