JSP+sql实验教学管理系统(系统+论文+开题报告+封面+中期检查表+英文文献).zip

Journal of Network and Computer Applications (2000) 23, 93–107

doi:10.1006/jnca.2000.0104, available online at http://www.idealibrary.com on

A simulated network management

information base

Colin Pattinson

School of Computing, Leeds Metropolitan University, Leeds LS6 3QS, UK,

e-mail:C.Pattinson@lmu.ac.uk

(Received 24 September 1999; accepted 17 January 2000)

Initially motivated by the requirement to provide trainee network managers with realistic

network whose behaviour, as represented by the variation in the values of data objects

within the management information base (MIB), is virtually indistinguishable from a real

network with real network entities.

This paper describes the processes by which these patterns are controlled, offers an

example of how the type of network behaviour reported in published measurements can

be simulated, and concludes by proposing the use of such an approach in the study of

the behaviour of network managers.

 2000 Academic Press

1. Introduction

The near-universal adoption of the computer network as a means to deliver

undetected faults, unsatisfactory levels of performance or breaches of security,

whilst the organization is unaware of the deployment or usage of the network

devices, and cannot determine appropriate mechanisms to recover costs from

users. As a consequence of this, a number of technological solutions have been

developed to assist the network manager in fulfilling some or all of the tasks

required. Typically based around the structures and data definitions of the Simple

Network Management Protocol (SNMP), tools are available which allow the

network manager to interrogate, control and study the behaviour of network

devices, with each device providing status information via a ‘management agent’,

1084–8045/00/020093 C 15 $35.00/0  2000 Academic Press

utilized in the work described in this paper, although it is envisaged that the

overall approach is transferable to other platforms which offer broadly the same

characteristics. The consequential benefits of this, in providing a common eval-

uation baseline, are described fully later.

The variety of possible tools and techniques which can be encountered

by the trainee computer network manager are discussed in a number of text

books: Rose [3] and Stallings [4] discuss the details of the protocols and data

structures used in NM, with an emphasis on SNMP; Leinwand and Conroy [1]

provide practical examples, primarily of ‘fault’ situations, which illustrate how

the information provided in an SNMP management information base (MIB) is

offer a controlled, repeatable, predictable (to the designer/observer, not the user)

and graduated set of tasks. This paper discusses the ways in which network

performance can be described statistically and identifies the way in which a

network can be modelled in a variety of simulation techniques. The paper then

describes a novel approach in which the techniques of simulation

are used to

control the data structures and variables of an artificial SNMP MIB, in such

a way as to permit the development of ‘model agents’ whose behaviour and

characteristics can be controlled to produce behavioural data identical to that

more appropriately; they can act as benchmarks for the assessment of different

hardware and software configurations; they can be used, in conjunction with

simulation/modelling, as part of the planning process. Statistical results relating

For the purposes of this paper, the distinction between the terms ‘model’ and ‘simulation’ is that

used by [1], p. 121, ‘The model is a set of rules the simulation needs to follow’, thus the simulated

network is defined by a set of model agents, each with its own model MIB.

published over the years [5–7], describing the characteristics of networks under

a variety of loading conditions.

These studies reveal that:

ž traffic flows are different for each end system [6] and vary over time [6, 7].

These patterns, and the fact that they can be described mathematically, imply

that a mathematical model can be developed which would accurately represent

the flow of data on the network.

3. Network simulation—related work

Another important tool for the network manager is simulation, which allows the

construction of (mathematical) representations of whole or parts of a network.

This then allows questions of the ‘what if’ form to be answered in support

of the planning process, or permits studies of the characteristics of particular

network design points without the need to obtain and install component parts.

of Service (QoS) in multimedia networks. Typical of this work is the way in

which Tse and Grosslauger develop a CAC process, and then model that process

to evaluate its effectiveness in delivering some QoS on a particular (simulated)

network [9, 10].

An example of the use of a ‘whole network’ simulation, in which users can

vary characteristics during operation, is MacDonald’s discrete event simulator,

cnet [11]. In cnet, protocol implementations are tested on a simulated wide

area network whose performance parameters (data loss, transfer rates, etc.) can

be varied in a controlled manner. The motivation here is to provide a tool

under which protocol implementations can be tested in a controlled environment.

systems: Kheradpir et al. [13] report on the use of simulation techniques to eval-

uate network management architectures in circuit switched networks. Perhaps of

tives are to validate the information model used within the TMN, and to provide

a reference configuration against which to evaluate NM tools. Whilst the major

motivation for the work described in this paper is to study NM practices and

methods, a similar approach could be taken here to develop reference configura-

tions for evaluating SNMP-based computer network management systems.

4. A statistical description of network behaviour

It is clear from the above that a variety of network types, applications and

behavioural patterns can be fully described by the use of an appropriate set of per-

formance measures (whether derived from measurement or simulation/modelling)

in combination with other data relating to the physical and logical configuration

so widening the range of data sets available.

The opportunity afforded by the use of the MIB’s data constructs allows

the method described here to be used in another important area of study, the

comparative evaluation of network management platforms themselves. The use of

a common set of network management data permits the same simulated network

to be ‘managed’ by any network management platform which conforms to the

SNMP specifications. This would allow a platform to be evaluated so as to

determine its effectiveness in a set of network management tasks, the evaluation

being either of an absolute nature (does the platform permit management on

the network under particular circumstances and requirements?) or a comparative