×

SUBSCRIBE TO TMCnet
TMCnet - World's Largest Communications and Technology Community

CHANNEL BY TOPICS


QUICK LINKS




 

Industry_Insight.gif (5056 bytes)
July 1999


IP Telephony Building Blocks

BY LAURENCE J. FROMM

The pace of change in the Internet industries is increasing. In an environment of rapid change, the winners are the companies that not only have the best products or services, but those that can most rapidly modify and improve their offerings. Speed is critical.

For a product company, then, fundamental architectural and design decisions — often made in the early under-funded, over-hurried days of a company — have long-term implications on the extensibility of the product family, and hence affect the ongoing competitiveness of the company.

IP telephony gateway vendors, for example, need to consider their view of the future in designing their products.

VIEWS OF THE WORLD
One view of the future is that IP telephony gateways are relatively simple, fixed-function devices. They convert telephony traffic to and from IP traffic, making a limited set of signal processing and protocol translations in the process. The best way to add value in this view of the world is to minimize cost and maximize performance. Adherents to this view handcraft software to minimize memory space for program code and data, and reduce procedure calls between software interfaces. They employ simple schedulers that efficiently dispatch a series of related tasks. They build embedded devices based on private architectures.

THE NEED FOR SPEED
Another view of the future is that IP telephony gateways sit on the edge of a new kind of network that enables new kinds of services. The gateways may be used to perform unknown and unknowable signal processing and protocol translations for some future services. Design for extensibility becomes another key criteria, along with cost and performance. Adherents to this view are more likely to define well architected interfaces between software modules; to use higher level programming languages more often; and to use real-time operating systems, enabling multiple tasks to execute within the same subsystem. They use industry standards at the interfaces to enable incorporation of third-party technology.

Gateways may be initially deployed doing a limited set of signal processing translations. Those built for extensibility can add capabilities to enable new services. The gateways may be extended, for example, to provision network-based interactive voice response (IVR), or for waveform capture for natural language speech recognition. Basic designs can be extended for video signal processing, or for multiparty, multimedia conferences.

DESIGN CONSIDERATIONS
Regardless of world view, gateway vendors consider a range of design choices in building gateways. Some of these include:

DSP Processor Architectures
DSP processor performance is characterized by various factors, including arithmetic format (fixed or floating point), instruction set type (CISC or RISC), clock speed, pipelining, memory organization, and peripherals.
There are two basic generations of DSP processors.

The current generation — represented by the TI C5000 and Motorola 563xx families of DSPs — are based on programmable filter architectures. These processors have a data path architecture targeted at DSP multiply accumulate operations. They incorporate a multiplier, an arithmetic logic unit, shifter, registers, and other specialized units.

The next generation of DSPs is based on RISC technology, with load/store architectures with multiple function units accessing large register files. Most, like the TI C6x family, are VLIW (Very Long Instruction Word)-based. Their orthogonal instruction set allows for optimized compiler implementations. The advantage of this architecture is that the processor can run at a much higher clock speed for increased performance.

Other Design Choices
Besides the DSP type and specific DSP, other DSP system design choices exist:
Memory organization: Design factors include the amount of internal memory, and whether it can be used as cache; whether to deployed local or shared external memory; and memory type (SRAM, DRAM, SDRAM).

I/O access: Multifunctional serial ports or hardware DMA are choices.
Run-time scheduling: Monolithic runtime environments are simple and limited. A program polls for data at fixed points, executes a fixed algorithm, and outputs data at the rate required. Priority stack schedulers are more powerful. These schedules support multiple contexts (or threads) with multiple priorities, but are limited to a single context stack and a single active thread per priority. Task-based real-time operating systems are the most general schedulers. RTOS support multiple contexts with multiple priorities, each with a completely separate context including a private stack.

THE BEST CHOICE
Comparing DSP processors solely on core functions like FFT or filtering can be misleading. The DSP system architecture needs to be factored into the decision process. The best performance benchmarking is done by comparing performance on the set of functions needed for a particular service or application. For example, an application that requires DSP processors to execute limited functions from internal memory is best matched with a processor and architecture with low power and footprint and adequate internal memory. A software-driven architecture, however, that supports multiple functions executing on a single processor for multiple channels, requires a powerful processor with high throughput memory and I/O interfaces, and excellent high-level language support.

If the new services enabled by converged networks require constant new capabilities from the edge devices, vendors that have a compatible world view win. If evolution of gateways, conversely, is characterized by bigger, faster, denser versions, then vendors with that world view win. Let the battle begin. c

Laurence J. Fromm is vice president, new business development for Dialogic Corporation. Dialogic is a leading manufacturer of high-performance, standards-based computer telephony components. Dialogic products are used in fax, data, voice recognition, speech synthesis, and call center management CT applications. The company is headquartered in Parsippany, New Jersey, with regional headquarters in Tokyo and Brussels, and sales offices worldwide. For more information, visit the Dialogic Web site at www.dialogic.com.







Technology Marketing Corporation

2 Trap Falls Road Suite 106, Shelton, CT 06484 USA
Ph: +1-203-852-6800, 800-243-6002

General comments: [email protected].
Comments about this site: [email protected].

STAY CURRENT YOUR WAY

© 2024 Technology Marketing Corporation. All rights reserved | Privacy Policy