Design Choices for Teledatacom Systems

With integration, consolidation, and networking of information (voice, video, and data) gaining momentum, engineers face new complexities and uncertainties. Learning to deal with the new technology starts with understanding the various types of nodes for what Artesyn calls the teledatacom infrastructure—as well as the protocols that will get data from node to node.

Today, there are essentially two main categories of transport networks: circuit-switching and packet-switching. A circuit-switching network establishes a physical circuit connection across the network before any data flows. The circuit is set up rapidly under computer control, and remains set up while the data passes, which might take a second or less. Then the circuit may be disconnected so that others can use the same facilities. If there are not sufficient network paths, some form of notification (i.e. a busy signal) prevents additional data flow. The strength (and weakness) of circuit-switched networks is that they provide a fixed-cost and fixed-bandwidth solution—independent of the traffic.

A packet-switching network divides the data traffic into blocks, called packets. Each packet of data travels in a data frame, or envelope, which gives the destination address of the packet and various types of control information. The packets eventually reach their destination where the envelope is stripped off—leaving the data. Such networks handle individual packets as they arrive and don't simply block new connections when they are overloaded. Packet-based networks mix traffic over a single circuit (and employ various techniques to compress voice signals). Thus they can use the network more efficiently when integrating voice and data—to provide a low cost, high-performance, shared bandwidth solution. Packet-switched networks are more cost-efficient. Why?

• They require no call set-up time (resulting in faster delivery of traffic).
• Users can efficiently share the same channel (resulting in lower cost).
• Packet switched networks permit session-less communications (resulting in the ability to push messages all the way through to recipient—as opposed to waiting for recipient to pull them through).

Advantages for packet networks over circuit switching are driving the convergence of the public switched telephone network (PSTN) and the Internet (or, more generally, networks using the Internet Protocol). First, packet switching provides the most efficient transport for multiple capabilities: voice, video, and data. Also, it is more efficient for bursty data traffic (which is growing ten times faster than voice).

Second, customers can save a lot of money on long distance calls and fax charges by transporting data and fax integrated in a single network with voice traffic. Third, multi-service support for several key packet-networking technologies, including Internet Protocol (IP), Asynchronous Transfer Mode (ATM), Integrated Services Data Network (ISDN), and Frame Relay, allows them to share the same network. Unfortunately, however, the non-deterministic behavior of packet-switched networks (such as the Internet) becomes a drawback for carrying voice.

Figure 1:  A dataflow diagram allows you to analyze the interactions, processing, and flow of data before designing the network.

Early in the design assessment, you must determine information flow patterns. Creating a data-flow diagram will greatly assist in this effort by keeping the design parameters focused. It also helps you visualize flow patterns-including the source, destination, transport medium, type and volume of data you can expect over the lifetime of the system you are designing. In addition, you may want to design in reserve capacity for traffic growth and for cut over (in case an element of the system goes down and you need to reroute data to another element).

With your data-flow diagram you can examine the performance options and needs of your system. For example, how will a given data stream be manipulated and routed? Will you have intelligent-access equipment with a data interface to a core ATM-based network? Or IP interfaces over ATM or Wave Division Multiplexing (WDM)? How will you deal with routing the timing-sensitive requirements of your data over a traditional public data route (such as the Internet) when you have little or no control over security and the traffic-congestion levels of its infrastructure?

The transition to an integrated teledatacom network requires new infrastructure employing a variety of interdependent technologies for interconnections. These sophisticated technologies (such as IP, ISDN, ATM, and Frame Relay) are optimized for data communications—with greater speed, performance, and bandwidth.

For more information on packet technologies:
IP • VoIP • Gatekeepers • ATM • Frame Relay

Which technologies and modulation schemes will come out on top for traveling the "last mile to the home"? Cable modems and wireless alternatives offer solutions, with inherent strengths for each. New broadband-access technologies, including an assortment of Digital Subscriber Line (DSL) schemes, could help alleviate the problem. But widespread deployment may be slow coming.

For more information on broadband technologies:
DSL • Cable Modems • ISDN

To achieve greater scalability and faster time to market, OEMs are moving to open platforms. Compared to proprietary systems, the use of standard buses or open system buses (such as VME, Compact PCI, and PMC) provides equipment manufacturers and system integrators with a variety of competitive advantages, including shorter development cycles, potentially lower costs, second sourcing, and simplified upgrades and maintenance. Suppliers and partners can also provide development experience and low-level support.

Figure 4:  An open-system based solution speeds development and lowers costs by standardizing the hardware and communications interfaces.

For more information on hardware platforms:
cPCI • VME • PMC

For more information on processors:
Motorola 68K • PowerPC • SPARC • Intel

In selecting the RTOS that meets your requirements, it's important to consider support for desired functionality/operability, such as legacy applications, processor support, intrinsic development tools/APIs, industry support/acceptability (in other words, will your OS mesh with rest of your system?), upgrade path, and longevity. Also there may reasons for mixing various processors to achieve your system requirements. For example, you may want a PowerPC processor for managing communication streams, while incorporating an Intel-based processor running Windows 2000 to meet graphical interface requirements.

Meanwhile, improved quality and reliability for PC-based systems (along with readily available PC-based software) is creating some pretty swift currents in the embedded OS market. Having long focused on the desktop, Microsoft is now addressing the real-time environment with its Windows 2000 and Windows CE operating systems.

For more information on real-time OSs:
Windows 2000 • Windows CE

Jeff Durst is a product manager for Artesyn Communications Products, a subsidiary of Artesyn Technologies, in Madison, WI. Jeff has been with the company for 13 years. Originally hired as a hardware engineer, he designed single-board computers used in military, scientific, and communications applications. Later, as a project manager, he supervised projects for several of today's major telecommunications companies. Now he concentrates his efforts on product definition.