Previous | Table of Contents | Next |
Roshan L. Sharma
This chapter provides valuable information on those critical elements of the network managers job network planning, design, and optimization. Included are descriptions of an effective network planning effort, modeling and performance issues, and tools for network design and optimization.
Network planning, design, and optimization are important components of the network management process. Traditionally, these functions have been performed through the use of powerful mainframe computers. Because these computers required the use of large tariff-related databases, a great deal of time was spent entering input data and interpreting the output data that invariably came in the form of a thick stack of computer printouts. No graphics were available to illustrate the network topologies. Furthermore, the user was always kept out of the design process. However, advances in very large scale integration (VLSI) technology have made powerful PCs available, which has opened the door to the development of better network planning and design tools.
The network planning and design effort can be broken into the following distinct tasks:
Creating the enterprise database is by far the most time-consuming of all network design tasks. An enterprise database (EDB) should at least list:
The list can grow into a very large one when the database must also classify the users at each location and their communications needs. However, the tasks involved in network planning, design, and optimization are impossible without the availability of an EDB. The table that appears later in this chapter illustrates a sample EDB.
The next three network planning tasks demand a capability for traffic modeling and analysis. Before defining the traffic engineering efforts, some basic traffic-related concepts should be introduced.
There are two types of traffic encountered in enterprise networks:
It is always assumed that connection-oriented voice traffic behaves in a predictable fashion, which implies that:
But a close observation of speech energy over the duration of a conversation will show that there are many pauses. Furthermore, two of the four-wire access lines (ALs) and trunks are always idle since only one party can talk at a time. These facts have helped long-distance carriers send more calls over expensive ocean cables than are possible over available trunks using pure circuit-switching by using the time-asynchronous speech interpolator (TASI) technology. Such a technology was never cost effective over cheaper land-based leased lines. With the availability of asynchronous transfer mode (ATM) and Broadband Integrated Services Digital Networks (B-ISDN), users can get the same benefit through the use of variable bit rate (VBR) capability.
The data traffic between two CPEs is always bursty because of the complex rules of data communication protocols. Very small control messages may be involved in both directions before user information can flow. Although a full-duplex connection can be maintained, shared transmission lines in a packet-switched network can carry variable-length packets from many sources concurrently, thus muddying the picture. The faster the transmission lines, the burstier the transmission will appear.
Circuit-switched voice and video traffic intensity is measured in erlangs, which is equal to the average number of circuits busy during a busy hour between two network nodes. For example, if 15.5 conversations are observed concurrently between two network nodes (e.g., between a PABX and a voice switch or over an access line bundle) during a busy hour, then the voice traffic intensity is 15.5 erlangs.
Previous | Table of Contents | Next |