Previous Table of Contents Next


PACKET-SWITCHED DATA TRAFFIC

Packet-switched data traffic intensity can be measured as the traffic rate in bits per second (bps) during a busy hour. Only the data rate in bps can describe the bursty nature of data traffic. Experienced network specialists have been using the concept of data erlangs for many years in defining the average data traffic intensity between two network nodes. This is obtained by dividing the observed busy hour data rate (R) by the capacity (C) of each separate transmission line. For example, if the busy hour data rate between two nodes is 392,000 bps and the capacity of a transmission line is 56,000 bps, then the data traffic intensity is 7 erlangs.

MODELING TRAFFIC FLOWS IN A BRAND NEW ENTERPRISE NETWORK

It is sometimes difficult to model traffic flows for a brand-new system. Many approximate methods have been devised for predicting traffic intensities (TIs) between all major CPEs. For example, a voice LAN (or PABX) generates about 0.1 * Ns erlangs of busy-hour traffic, where Ns is the number of active subscribers served by the PABX.

A breakdown of these traffic expressions into intranodal and internodal traffic should be determined by the known pattern observed at each enterprise. Some network designers use the 70/30 breakdown — 70% of the traffic remains within the site (voice/data LAN) and 30% of the traffic goes to other CPEs as internodal flows. These TI values can then be entered into an input file that defines each site ID, the related vertical and horizontal coordinates, and the total traffic intensity handled by the site.

The next task is to model the internodal traffic flows (i.e., exact traffic intensities handled by all the nodes and links in the path of a CPE-CPE connection). These computations are generally performed by the network design software for each assumed network topology (i.e., number of network switches and the link types employed at each network hierarchy). Some tools use critical design parameters to determine the fraction of traffic handled by access lines (connecting CPE and a switch) and trunks (connecting two switches). Eventually, the tool provides the total traffic intensity handled by each resource (node or link) of each network topology considered during a typical busy hour.

MODELING TRAFFIC FLOWS IN AN EXISTING ENTERPRISE NETWORK

Exact traffic flows can be modeled using the detailed traffic data gathered by intelligent network nodes (e.g., PABX or LAN). The source ID, destination ID, call originating time, and call duration for each connection is recorded in station message data recording (SMDR) tapes of the voice network. Similar data is recorded by the data LAN for the packetized traffic. Simple traffic analysis packages are obtainable for analyzing the exact internodal traffic patterns between all pairs of CPEs. Such data can then be entered in a from-to data file (FTF) to define CPE traffic as simple vectors (i.e., From-Node ID, To-Node ID, and the BHR traffic intensity) for each CPE-nodal pair.

This effort eventually provides actual traffic flows (i.e., the actual traffic intensity handled by all resource, nodes, and links) of each network topology studied during a typical busy hour.

MODELING TIME-CONSISTENT AVERAGES (TCAS) OF TRAFFIC FLOWS

Choosing a “busy” hour is an important task. Networks are not cost effective when they are operating during the hour with the highest traffic. A network may provide the required grade-of-service (GOS) during the busiest hour, but at all other hours of the day (especially during the evening and night hours), the GOS level would be overkill. No organization can afford such a network. Network managers who select an hour with the least traffic during the day will hear complaints all day long. Therefore, a proven methodology is needed to select the average traffic intensity for network design. There are two methodologies — one used in North America and one used in all other countries.

The first methodology requires the selection of a typical month and the creation of a matrix (30 ¥ 24) of traffic intensities (TIs) for each network resource for that month. Next, the average traffic intensity for each hour of the day over all 30 days of the month is computed. This process is repeated for each hour of the next 24. The TCA traffic is the maximum value of all 24 TCA values. This value determines the size of the resource (i.e., number of AL and trunks in the bundle connecting two nodes or the computing power of an intelligent node). It is helpful to have a software package for computing TCA traffic intensity (TI) values.

The second methodology requires that the 36 highest TI values be observed over an entire year and then the average computed to get a TCA value. This must be done for all resources.

Both of these methodologies result in more economical networks. However, no single methodology can predict an exact traffic pattern. Traffic values behave like the stock market. A single catastrophe, such as an earthquake, can also change the traffic patterns drastically. The objective of an effective traffic engineering practice is to synthesize an economical enterprise network using a consistent approach.

MODELING TRAFFIC GROWTH DURING THE SYSTEM LIFE CYCLE

To estimate the total costs incurred during the life cycle of a network system, the traffic intensities for each year of the life cycle should be modeled. The Delphi approach often works best. In this method, all general managers are interviewed and realistic models of traffic growth during every year of the life cycle can be built. Some divisions may disappear through divestiture or attrition. The data from all of the interviews must be collected, weighed, and processed to create a meaningful model.

PERFORMANCE ISSUES

Before performance requirements for all the communication needs of an enterprise can be defined, network managers must first study the available concepts of performance and identify the exact enterprise needs in each business area.


Previous Table of Contents Next

Copyright © CRC Press LLC