Handbook of Local Area Networks, 1998 Edition:LAN-based Application Development Issues and Solutions Click Here! Search the site:   ITLibrary ITKnowledge EXPERT SEARCH Programming Languages Databases Security Web Services Network Services Middleware Components Operating Systems User Interfaces Groupware & Collaboration Content Management Productivity Applications Hardware Fun & Games EarthWeb sites Crossnodes Datamation Developer.com DICE EarthWeb.com EarthWeb Direct ERP Hub Gamelan GoCertify.com HTMLGoodies Intranet Journal IT Knowledge IT Library JavaGoodies JARS JavaScripts.com open source IT RoadCoders Y2K Info Previous Table of Contents Next Step One: Establishing End-User Needs This step, paradoxically, is the most difficult one in many companies. End users do not speak technical language and often do not know exactly how to translate their terms. However, without understanding exactly what the end-user needs are, any modeling or simulation process will be of little value. Ultimately, the simulation model predicts the end-user response time of the client/server application. To establish needs, there must be meetings between end users and applications developers. Quantitative analysis of the number of users, operating systems, machine types, new applications, and required response time must be carried out. A side benefit of this process is that it encourages and instills a team approach to IS infrastructure that involves all interested parties. Step Two: Benchmarking Existing Traffic Many firms have implemented centralized network management systems as well as distributed remote monitoring (RMON) probes over the last three years. Network topology information can be captured from the management platform and input to the simulation tool directly. In addition, RMON probes—along with portable analyzers—permit the characterization of the existing traffic on local area and wide area network segments. Traffic information can be input into simulation models directly as a captured packet trace. Simulation tools such as COMNET III (CACI Products Corp., LaJolla CA) have interfaces For Network General Sniffer traces as well as traces from Hewlett-Packard’s NetMetrix RMON agents. Using COMNET III, it is also possible to import topology information from HP’s OpenView, IBM NetView 6000, and Cabletron Spectrum. Step Three: Characterizing New Applications Most simulation projects are motivated by the addition or expanded use of a particular application. To characterize a new application, it is necessary to have access to that application, preferably on a production network. A test operator runs the application and network analyzers capture the traces. It is important to understand the operation of the application to create an accurate model. Use of statistical techniques alone can be dangerous and lead to erroneous results. Following the characterization of a single user, a small group of three to five test users should be monitored. Based on the traffic from this group, a model can be created for an individual user. The model needs to take into account the statistics of the source and include such parameters as peak and average bytes-per-second, peak and average frames-per-second, and spacing of peaks in time. Time resolution for these measurements should be one second. Step Four: Creating and Running the Simulation Model At this point, all information has been collected to create and run a simulation model. This model incorporates the network topology and background traffic from steps one and two and the models for new applications from step three. Quality-of-service metrics from step one can be used to evaluate the simulation runs. Quantitative Results Simulations aim to exactly duplicate the packet conditions on the wire. They include frame headers and incorporate standard routing algorithms such as remote image processing (RIP) and interior gateway routing protocol (IGRP). They include transport protocols, such as Transmission Control Protocol (TCP) and sequenced packet exchange (SPX). Simulation engines such as CACI’s COMNET III also include client and server icons that can be customized by the user. Often, application bottlenecks are in the server and not the network itself. Simulation models are selected to evaluate WAN link capacity, server performance, or LAN technologies. Each run takes from minutes to hours, depending on the model complexity and machine horsepower. In a matter of a few days it is possible to generate a series of simulations that can definitively evaluate a network application problem. Quantitative results are obtained on network utilization, router buffer utilization and headroom, statistical distribution or application response times, dropped packets, effects of prioritization on response time, effects of transport window size on throughput, and many other variables. The following examples describe projects where simulation has been used to answer difficult networking questions. Results were obtained using HP’s Network Advisor, NetMetrix RMON probes for data collection, and CACI’s CONMET III for simulation. CASE STUDIES Market Data Distribution A major bank is planning a new, customized market data distribution system to be supplied over a backbone network infrastructure to 1,500 potential users. The market data server will be centrally located on an Ethernet LAN. The server takes digital line fees from the market data vendor and addresses the information to client subscribers. Such market data services include stock quotes, spreadsheet analyses of trends, and graphical representations of financial information—all of which can generate significant network traffic. Tests were run with one user on the bank’s production network. From these tests, a model was created that allowed simulation of the traffic generated by more users. A simulation run modeling 100 users with three profiles open is shown in Exhibit 6-5-2. Exhibit 6-5-2.  Simulated Traffic from 100 Users of Market Data Application—Ethernet Segment Peak utilization was determined to be 47% of the 10M-bps capacity of Ethernet. These results illustrated that the application would not fit on the Ethernet server backbone. In this case, the application was rewritten to reduce overhead traffic. The application is now successfully running. During the initial deployment of the first version of the application, further simulation models were run and compared with actual network traffic. This step—model validation—is an essential component of any simulation project. Traffic measurements of a small group of five users were made using an HP Network Advisor protocol analyzer. Simulation runs were carried out to model this group. The simulated traffic replicated the measured traffic within 5% to 7%. Previous Table of Contents Next Use of this site is subject certain Terms & Conditions. Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details.