Čńňî÷íčę: http://www.cms.livjm.ac.uk/pgnet2007/Proceedings/Papers/2007-011.pdf

Network Modelling and Simulation Tools


Muhammad Azizur Rahman1, Algirdas Pakstas1, Frank Zhigang Wang2
1London Metropolitan University, 66-220 Holloway Road, London N7 8DB England
2Cranfield University, Cranfield, Bedfordshire MK43 0AL, England
1mam406@londonmet.ac.uk, 1a.pakstas@londonmet.ac.uk, 2f.wang@Cranfield.ac.uk




ABSTRACT

  Computer network technologies have been growing explosively and the study in computer networks is being a challenging task. To make this task easy, different users, researchers and companies have developed different network modelling and simulation (MS) tools. These network MS tools can be used in education and research as well as practical purposes. They vary with their characteristics. This paper reviews some of the most important network MS tools developed recently. This paper also shows a classification of the tools used in communications networks.

I. INTRODUCTION

  Network is a complex mix of applications, communications protocols and link technologies, traffic flows and routing algorithms. It is immensely complex. Network design process is a challenging task, requiring designers to balance user performance expectations with costs and capacities. One of the obvious approaches to deal with complexity is the use of modelling and simulation (MS) techniques. There are currently available various MS tools created by the separate companies and groups of researchers in academia which are intended for use as practical and/or educational tools for network design. There is a need to study the existing network MS tools, their functionality, advantages and function procedure and special features. Nearly every general-purpose network design tool works the same way [1], sometimes having complementary features. This purpose of this paper is to review these MS tools

  Rest of the paper is organised as follows. Section II describes the classification of MS tools. Section III and Section IV describe the most used and recently developed MS tools. Finally, Section IV provides conclusions.

II. CLASSIFICATION OF THE NETWORK MS TOOLS

  There are different types of network MS tools available and they are used in many ways. According to the uses and availability (we analysed 75 tools), the network tools can be classified into four groups: analytical tools, simulation tools, topology discovery tools and topology generation tools. Analytical tools help to design a network model and calculate different factors (e.g. reliability, utilisation) of the model. On the contrary, simulation tools simulate dynamic behaviors (e.g. packet passing, link failure, TCP protocol) of a network model besides modelling. The topology generation tools help to generate small as well as large topologies based on different algorithms. Finally, the topology discovery tools extract the actual network information from an existing system and map them graphically and/or in text format. This type of tool is used for network management purposes. Both analytical tool and topology discovery tool can be classified into two: educational (free) and commercial tools. Network simulation tools can be divided into three: educational (free), commercial and specialised tool. The Fig. 1 shows the classification of the tools. This paper is focusing on only the analytical and simulation tools.

III. ANALYTICAL TOOLS FOR NETWORK DESIGN

  A. Educational Tools

  DEsign tool LITE, Delite [2] is an educational and practical wide area network (WAN) design tool, which can produce network designs of limited size using a set of the embedded network design algorithms. Delite can produce graphical displays representing network nodes and links as well as some additional analysis data (delay analysis, reliability analysis, average delay analysis of the link, average number of hops, link utilisation for every separate link between nodes, utilisation of each node, overall network model utilisation, etc.). There are seven files that are handled by the Delite tool for each network design. However, the most important for the users are .gen (original node information, coordinates and available link types) and .net (additionally has table of links between nodes i.e. actual design) files. Links between nodes can be generated using a few design algorithms. Thus, various designs may have different costs, delays, reliability, average number of hops etc. Other files are used for different purposes: noeqip.tbl and pararatbl are predefined (used to supply some network parameters), .cst file describes the costs associated with link types (e.g. Tl, T3, D96 etc.), .req file describes capacity of each link type, .inp file contains the names of files related to a particular model. The users can edit all of the files mentioned above as they are in ASCII text format. The algorithms used for designing the topology of a network model supported by Delite are: Prim, Primdijkstra, Tour (Nearest Neighbour), Tour (Farthest Neighbour), Esau-Williams, Sharma, Multispeed Tree Design, Nearest Neighbour-EsauWilliams, Multicenter Esau_William, Mentor, Incrementour. Contributor: Robert S. Cahn. Supports: different types of network design algoriths (mentioned above). Languages: C. Advantages: 1) Text formated input and output files so user can manually change the input files to see the effects; 2) Population is taking part in designing the models; 3) World map can be included in the model to see the suitability of the model; 4) can include more algorithms in this tool. Limitation: 1) limited number of nodes supported as this is a small scale network tool; 2) documentatoin is no rich enough.

Classification of the Network Design and Simulation tools

Fig. 1: Classification of the Network Design and Simulation tools

  Cappuccino [3] is a web based non-commercial network design tool. Using the tool, users can design network model, calculate maximum traffic of each link and a lower bound of the link cost of the network. The switches can be scattered randomly inside the rectangular area (0, 0, 400, 400). The properties of the switches are expressed by switch name, X coordinate, Y coordinate, Alpha (originating capacity for a switch) and Omega (termination capacity for a switch), which can be changed by the users. The Alpha denotes the maximum communication rate, which is permitted by the network to originate from switch to the rest of the network. The Omega denotes the maximum communication rate, which is permitted by the network to terminate to a switch from the rest of the network. Traffic between switches is limited by using set-pair-constraints, which specifies the maximum traffic between two switch sets. While designing a network model, initially 20 switches are scattered in a region. User can add, delete and modify the switches as required. Topology is chosen using predefined algorithms implemented in the tool or manually. Then traffic constraint, which specifies maximum traffic between two sets of switches, is chosen. The cost coefficient is proportional to the link length and a link has two-bandwidth capacity (incoming and outgoing). The tool uses shortest path routing algorithm for traffic. The algorithms for network topology implemented in the tool are: best star, minimum panning tree (Krustral's algorithm), delaunay triangulation and complete graph. Using shortest path routing and taking account the constraints, the tool calculates the maximum traffic and a lower bound cost on every link so that user can see how far away the network is from the optimal. The lower bound cost of the link is calculated based on complete graph. Users can also analyse the traffic of each link, switch and traffic constraints of each using window menu. User can edit the switch information and update the calculation. The tool is very user friendly. It is platform independent. This tool will help network users/designers create their non-blocking network model. Languages: Java. Advantages: This is web based. Limitation: Very limited applications.

  B. Commercial Tools

  XNetMod [4, 5] (based on NetMod tool), a LAN modelling tool that allows users to analyse the performance of a configuration before implementing it. Three approaches to performance evaluation of this tool are analysis, simulation and measurements. XNetMod can be used networking environment consisting of thousands of computers sites. XNetMod provides customised forms to enable one to specify input parameters for the elements. XNetMod supports two types of topology: ring and bus and two connectors: Router and Bridge. To facilitate the concept of bottom-up design, XNetMod allows users to cerate sunbnetworks. At present there is only one element type, Subnet in the Subnetwork element group. XNetMod supports analytic and simulation techniques. It can summarise the entire results of an analysis or simulation of just the result for a particular node. Supports: FDDI, token ring, token Bus, generic bridge, generic router, subneting. Languages: C. Advantages: 1) ggraphically display and manipulate network topology; 2) ease of subnetwork definition; 3) three types of analysis of data are supported; 4) flexibility of easily add or delete subnetworks; 5) ability to provide substantial user interaction for either simple parameter change of major reconfiguration of LANs and backbone; 6) hierarchical modelling capability for extremely large models; 7) query and generate reports based on currently displayed model.

  TND-Tool (Topological Network Design Tool) [6] is used to find out an effective solution to the network topology design problem. The idea behind developing the TND-Tool are twofold: first providing network designer and operator with a tool to facilitate their work, second offering topology design researchers an environment for testing and representing the output of their network design algorithms. It permits to select between node types, link types and data bases for different kinds of networks. A Network viewer allows to display the current topological network structure during the optimisation process. Users, engineers are able to set up a network interactively by inserting nodes and links, typing in network information, e.g. traffic values, moving, adding and removing some part of the network. According to the tariffs defined, the TND-Tool calculates the cost for an existing network on the screen or a network resulting from an optimisation algorithm. The TND-Tool consists of four major modules: 1) representation module is the graphical user interface of the system providing all important facilities needed by the network engineer; 2) optimisation module enables integration of network optimisation algorithms; 3) cost calculation module is the entity providing the rules for calculation of network cost, e.g. tariff tables and cost functions; 4) data base module provides all the needed data for the functionality of the system. Currently, the TND-Tool is used for designing GSM (Global System for Mobile communication) and DCS 1800 (Digital Cellular System 1800) fixed networks as part of the cellular mobile communication networks. The TND-Tool is not limited to cellular networks and has been designed as a platform for designing all kinds of networks such as Wireless Local Loop, Satellite networks, ATM networks and so on. Supports: WAN, Wireless, ATN, Satellite. Advantages: 1) an interactive menu and mouse-driven tool; 2) Web based. Limitation: 1) Limited only to calculate the cost of design; 2) Limited number of topoloy design algorigm supported

  NetRule [7] is designed to combine the best features of the mathematical analysis and simulation tool. It is mainly performance prediction tool to design the actual network. NetRule can model and stress test very large networks (1,000 to 10,000 nodes). It has a customised GUI like a simulator. It uses closed-form mathematical analysis that evaluates network load and performance in seconds with detail and accuracy that matches the best practice of mathematical analysis. The tool lets users to create or open network models and libraries, edit them graphically, and evaluate their performance. The diagram can animate message flow, highlight bottlenecks and shows utilisation and delays bars as components. It uses shared objects definitions to reduce modelling time and allow global changes. The NetRule library includes: 1) LANs, including many CSMA/CD, token ring and polling protocols; 2) Frame relay clouds with access ports and virtual circuits; 3) ATM switches and SONET links; 4) Routers, hubs, bridges and Ethernet switches; 5) Point-to-point Tl and other speeds; and 6) FDDI; 7) ISDN. Reports (output) are available for object definitions as well as for evaluation statistics that can be sorted or can be exported to spreadsheet, database, presentation or graphics packages. The most interesting features of the NetRule are period objects, which let one to establish different workloads during different times of day (day and night period) or time of years. Histogram can be produced for performance analysis. Normally the report includes: computer delay/job, computer utilisation, job cost, job delay, link delay/pkt, link utilisation. Another important feature of NetRule is that it can import network topology that has been discovered with a network management tool. Contributor: Analytical Engines Inc. Supports: CSMA/CD, Frame relay, ISDN, FDDI, ATM, SONET, Point to point Tl connection. Languages: Java. Advantages: 1) supports very large scale network (1,000 to 10,000 nodes); 2) user can import ALL or any part of a network and its parameters to build or modify a NetRule model. E.g .cap files, .csv format file,MIB data; 3) Supports generalisation option for clients, servers, and/or applications (for example, mathematically determining the typical client); 4) it can extract from traffic files and make a new file for all computers which can potentially be servers, all applications (using port numbers), user traffic flow.

  XNP (Extensible Network Planner) [8] is a comprehensive network design tool for both extensible and conventional networks. It is implemented using several methods and incorporated them into a previously unpublished java-based software tool, Cappuccino. Its objective is to configure a least-cost network that accommodates any traffic pattern satisfying the given traffic constrains. XNP allows network designers to quickly create, configure and evaluate network designs by providing a convenient graphic-based interface and automated functions. There are five essential building blocks in designing a network in XNP. Users can create a trail topology in the blank drawing area of the XNP main window. XNP allows two types of links (unidirectional and bidirectional). Application format (number of processing steps, number of machine instructions and bandwidth required) can be added or removed by user-friendly buttons. XNP provides three ways to specify traffic expectations. XNP process will iterate through all links and processing nodes to compute the capabilities for them. XNP will redraw the topology to show the differences in resources capabilities again and again. The thickness of each links shows its capability relative to other link’s capabilities. Users can evaluate the network configuration by computing a lower bound on the cost of the best network configuration. By adjusting options of lower bound and therefore by considering design restrictions for using resource again and again, one can obtain more accurate and useful measure. XNP allows three different ways to view the drawing area-only the trail topology and the design space together or the design space only. XNP also allows to manage multiple designs at a time so than one can try for different trial topologies and compare them. Supports: supported algorithms are: 1) Complete network; 2) Delaunay triangulation; 3) Delaunay triangulation with trimmed links; 4) Link complement; 4) Minimum spanning tree; 5) Random link adder; 6) Random node adder; 7) Star network; and 8) Symmetric link adder; 9) Geographical planar projection- longitude and latitude; 10) Grid network – N x M grid network; 11) Torus network- N x M torus network. 12) Localised traffic constraints; 13) Proportioned Pairwise traffic constraints. Languages: Java. Advantages: 1) Extensible.

  Table 1 shows the summary of analytical MS tools.

Table 1:Summary of network analytical tools
Tools Usages Design Algorithm(s) Target Network Model Outputs Stored Data File Platform
Delite Educational Some predefined algorithms WAN Graphical
Link analysis, node utilisation, overall utilisation in percent, delay analysis		 Special formatted Text file Windows
Cappuccino Educational Predefined Design algorithms WAN Link cost, capacity etc Graphical output, no text file Windows
Web based
XNetMod Commercial User defined topology LAN, Subnetwork Delay, average packet queue length, packet delay time, utilisation, Graphical output, No text file Windows
Unix
TND-Tool Commercial User defined topology WAN, Wireless, ATN, Satellite Analysis of Cost of deigned network Graphical output
No Text file		 Windows
NetRule Commercial User defined topology WAN, LAN, ATM switches, ISDN, FDDI Computer Delay, Computer Utilisation, Jobs cost, Job delay, Link Delay, Link Utilisation, histogram representation. Report can be imported to spreadsheet, database Windows
XNP Commercial Predefined design algorithms WAN Resource capacities, Cost analysis of links XNP formatted text file Windows
UNIX


IV. Network Simulation Tools

  A. Educational Tools

  ns-2 (Network Simulator) [9, 10], a VINT (Virtual Inter-Network Testbed) project from U.C. Berkeley/LBL/Xerox PARC, is a discrete event simulator targeted at network research, which provides substantial support for simulation of TCP, routing and multicast protocols. The simulator is using a Tcl/Otcl (Tool Command Language/Object Oriented Tel) as a command and configuration interface. There are four types of files related to ns-2 simulator. Model is described in files .tcl or .ns which have common subset of commands but not exactly compatible between each other's. While simulator runs a model defined in .tcl/.ns file, simulation trace file (.tr) and animation file (.nam) are created during the session. Network Animator (.nam) files are used to visualise the behaviour of the network protocols and traffic of the model. Once created, users can play with the .nam file just like a media player and check the behaviour repeatedly, ns-2 facilitates the three broad themes of network research simulations: 1) selecting a mechanism from several options; 2) exploring complex behaviour; and 3) investigating unforeseen multiple-protocol interaction. Supports: TCP family, UDP, CBR, FTP, HTTP, Pareto, Exponential protocols, wires, wireless, unicast, multicast. Languages: both C++ and OTcl languages. Outputs: dynamic output. Advantages: 1) user can design model both manually or writing code; 2) easy configurable and fast simulation by using two different languages (OTcl and C++); 3) many protocol already implemented; 5) dynamic behaviour can be visualised using nam editor; 6) Open source code and can be extended for future. Limitation: 1) Long time to get used to usign it; 2) does not large scale simuaton (e.g., internet). 3) incomplete API; 4) no real time simulation; 5) no performance analysis is possible 6) badly documented source code; 7) no direct support of mobility and shared wireless radio channels; 8) unable to scale to networks of Internet size due to the computational requiremets of fine-grain packet-level simulation and memory needed to maintain queues at each network link.

  Network Workbench [11] is a discrete event network simulator developed for the academic investigation of Internet protocol. It contains a complete protocol stack, abstracted from the Internet stack and a set of exercise that focus on critical protocol algorithms in the Internet stack. Network Workbench has several version, which are: version 0(1994), version 1(1994), version 2(1997), and version 3(1998). This tool supports several network topology, DLC error control, CSMA/CD collision backoff, optimal route computation, reliable transport, multicast, and LAN/MAN integration. It supports FDDI (Fiber Distributed Data Interface), OSPF (Open Shortest-Path-First). Workbench is abstracted from the internet (TCP/IP) stack. It contains five layers (application, transport, network, datalink control, and physical). The Network Workbench tool can be used to study of WAN topology, DLC (datalink control layer) frame formatting, DLC flow and error control, CSMA/CD local area network, network layer routing, reliable transport, multicast networking etc. Summary of statistics of the performance is produced at the end of the simulation. Contributor: J. Mark Pullen. Supports: network topology, DLC error control, CSMA/CD collision backoff, optimal route computation, reliable transport, multicast, and LAN/MAN integration, OSPF, FDDI, Ethernet, TCP/IP. Languages: C++. Limitation: Very limited implementation of protocols.

  Netsim (M.I.T.'s Network Simulator) [12] is a single process discrete event simulator used for the investigation of many aspects of Local Area Network (LAN). Netsim has three main goals: 1) flexibility of experiment of specifying the network and traffic; 2) simulate the accurate behaviour of Ethernet; and 3) has features that make running sequences of related experiments easier. The experimental data of Netsim simulation is stored in an experiment description file which contains: the layout of the network to be simulated, the traffic generation behaviour of the station on the network, information about repetitive of each run and about the sequence of parameters to be used for a series of experiments, etc. The entries specify the physical characteristics (e.g., length, packet size, data rate, number of attached stations) and traffic generation behaviour of the attached station. There are five types of traffic distributions in Netsim: exponential, uniform, deterministic, continuously queued and user defined discrete distribution (e.g. biomodal). Netsim collects information on many aspects of the operation of the simulated network. For example, average packet delay, histogram of packet queuing delays and collisions per packet transmitted, overall throughput (percentage utilisation), actual data rate, average, queuing delay, variance of queuing delay, observed packet transmission rate, average message size, total observed collision rate and total number of packet, maximum packet size, minimum packet size, maximum data rate etc. Contributor: MIT LCS Advanced Network Architecture group. Supports: Ethernet link, a point-to-point link, a switch (switches packets between several links), a host (about the same as a switch), Purdue's implementation of TCP, data supplier and consumer from TCP, a simple Poisson traffic source and a packet sink. Languages: C. Advantages: 1) Netsim is a publicly available ATM simulator originally developed at MIT; 2) its source code is freely available and modifiable; 3) Netsim has a user friendly Graphical Interface (GUI). Limitation: 1) The GUI however is very primitive and is inadequate for instructional purposes. It is not fault-tolerant to novice users. An incorrect sequence of keys or mouse clicks could cause the simulation to crash; 2) Its uses are limited.

  MaRS (Maryland Routing Simulator) [13, 14] is a discrete-event simulator proving a flexible platform for the evaluation and comparison of network routing algorithms. MaRS allows physical network, routing algorithm and traffic sources. MaRS is structured in two parts: a simulation engine, which manages the event list and user interface and a set of components for modelling the network configuration and handling certain simulation functions. MaRS has been used to evaluate and compare several next-hop routing algorithms: two distance-vector algorithms- a) Loop-free Bellman-Ford Routing Protocol Without Bouncing Effect, a Failsafe Distributed routing Protocol; b) a link-state algorithm (the New Routing Algorithm) for ARPANET. MaRS is limited due to the lack of scrolling capability in the user interface window. Thus, users are limited in what can be displayed to the size of the non-scrolling window. In addition, MaRS constrains itself to a point-to-point network structure and therefore is not readily suitable for simulating networks with broadcast communication. Contributor: University of Maryland. Supports: network routing protocols (SPF, MerlinSegall, BellmanFord), FTP, Telnet, Workload (Poison, uniform). Languages: C. Limitation: 1) Its support for transport layer protocol and application source models is very limited; 2) a primitive GUI is provided; 3) not available now.

  pdns (Parallel/Distributed ns) [15] simulator consists of extensions to the widely used and publicly available ns network simulator. The extensions allow many existing ns simulations to be run in a distributed environment with minimal changes. The pdns implementation also takes advantages of the large body of existing models found in and uses those without modification. Also includes in the pdns extensions is novel packet routing method called NIx-vector that allow routing decisions without the necessity of routing tables, resulting in substantial memory saving. While distributing the ns tool to several processors, for connecting each sub-model, IP address and a network mask is added to physical end point (nodes) for communication in logic as well as physical connection. For routing, each simulated node will start with routine table priori and dynamic routing information in the simulation time using Boarder Gateway Protocol (BGP) to adapt to any change in the simulated network topology. For event time management and event distribution among the parallel simulation sub-models, RTIKIY and LBTS (lower bound time-stamp) etc. time management run-time library are used. For synchronous event communication, multicast group management (MCAST) strategy is used. Contributor: The PADS research group at Georgia Institute of Technology. Advantages: 1) Very large-scale simulator (hundreds of thousands of nodes); 2) IP Addresses are supported.

  B. Commercial Tools

  OPNET (Optimised Network Engineering tool) [16, 17] was launched in 1987 as a first commercial available simulation tool for communication networks. It provides a comprehensive development environment for the specification, simulation and performance analysis of communication network. It can simulate all kind of wired networks and a 802.11 compliant MAC layer implementation is also provided. A large number of communication systems from a single LAN to global satellite networks can be supported. The most important features of OPNET are: modelling and simulation cycle (to assist user to go through three phases in design circle-building model, execution of simulation and the analysis of output), hierarchical modelling (describes different aspects of the complete model being simulated), specialised in communication networks (support for existing protocols and allow users to either modify theses existing models or develop new models), automatic simulation generation. An OPNET model consists of three layers: the network models, the node model and the process model. The tool allows the different layers in a protocol stack running on individual nodes to be represented. OPNET contains various tools for data collection. A probe editor allows specifying: 1) which statistics are to be collected from where (Probe editor); 2) their own statistics (analyse tool); 3) animation view and formats (Filter Tool, Animation viewer). OPNET contains objects that are capable of generation vast amount of output data during simulation. It can generate error rate and throughputs, delay queue size. Packet trace may be done. Output can be plotted in graph, such as end-to end delay vs. queue buffer capacity, loss ratio vs. queue buffer capacity. Probability distribution function, cumulative distribution function as well as histogram can be plotted for several data sets. OPNET is extensively used for the study of TCP transport across different types of ATM bearer capabilities and diffserv per hop behaviour. Contributor: Mil 3 Inc. Supports: routing protocols (OSPF, RIP, EIGRP, BGP, IGRP, DSR, TORA IS-IS, PNNI), Diffserv, MAC, mobility of nodes, ad hoc connectivity, different application models. Node failure models, modelling of power-consumptiion etc. Languages: C, Java. Advantages: 1) extendible; 2) large customer base; 3) professional support; 4) very well documented; 5) ships with a large number of built-in protocols. Limitation: 1) relatively high price; 2) complex, takes time to learn; 3) there is restriction to its portability.

  COMNET III [18, 19, 20] is a commercial integrated discrete event object-oriented simulator for modelling and performance analysis of computer network. With COMNET III, users can create a variety of network architectures, including LANs, MANs WANs, packet switching, ATM, frame relay and so on. COMNET is entirely driven by graphical user interface. Users are able to graphically display traffic modelling patterns, use of realistic network objects to reflect real networks and apply the network concepts. COMNET III allows the users to model, tune and analyse the performance of various types of networks. It provides an extensive library of nodes, links, protocols and traffic objects. The library of link objects include two classes: point to point links and multi-access links. Multi-access protocols that can be modelled include CSMA/CD, CSMA, ALOHA, Token BUS, Token Ring, FDDI and Polling. It supports subnets. It has two kinds of sources: application sources and traffic sources. There are four types of traffic generators: message sources, session sources, response sources and call sources. COMNET III is able to produce about 100 reports for different model building blocks. Most commonly produced reports include node utilisation and application delays, link delays, channel utilisation, message delays, packet delays, calls blocked, disconnected and preempted, session set up delay, collusion statistics, token ring statistics, buffer statistics etc. Contributor: CACI Company. Supports: LAN (Ethernet, Token ring, FDDI), MAN, internet, packet switch, circuit switch, ATM, bursty traffic, CSMA, CSMA/CD, ALOHA, Polling, Protocols (TCP/IP, IPX, SNA, DECNet). Advantages: 1) a graphical package, allow quickly and easily analyse and predict the performance of network; 2) allows users the flexibility to try an unlimited number of "what if scenarios; 3) realistic and accurate results. Limitation: 1) source code is not available; 2) new modules were very difficult to add; 3) It is restricted to experimenting with the set of networking protocols provided by the package; 4) It is not programmable by the users.

  REAL (REalistic And Large) [21] is a network simulator originally intended for studying the dynamic behaviour of flow and congestion control schemes in packet-switched data networks. It provides around 30 modules that exactly emulate the actions of several well-known flow control protocols (such as TCP), and 5 research scheduling disciplines (such as Fair Queuing and Hierarchical Round Robin). There are nearly 30 source types, corresponding to 30 or so transport protocol and workload types. The sources can be categorised into one of two types: flow-controlled and non-flow-controlled data sources. REAL has a graphical user interface (GUI) written in Java. The simulator takes as input a network scenario (a description of network topology, protocols, workload and control parameters) described in NetLanguage (ASCII). It supports sources or sinks node, gateways (synonymously with routers, bridges and switches). Users have to specify some network parameters, the transport protocol (in particular, the flow control) and the workload at each source. Finally, users must specify control parameters such as the latency and bandwidth of each communication line, the size of trunk board buffers, packet sizes etc. Contributor: S. Keshav at Cornell University. Supports: TCP/IP, XNS, FTP, Telnet, ill behaved, FIFO, FCFS, Fair Queuing (FQ), DEC congestion avoidance and Hierarchical Round Robin. Languages: C, JAVA (GUI only). Advantages: 1) the GUI allows users to quickly build simulation scenarios with a point-and-click interface; 2) source code is provided so that interested users can modify the simulator to their own purposes; 3) extendible. Limitation: 1) NEST (initial tool) didn't not allow for timers, REAL sends out a timer packet from a source back to itself to return after some specified time, but timers cannot be reset using this method; 2) misses some of the behaviours in the common protocol implementation; 3) do not support common implementation of protocols (direct execution). Instead it uses codes that simulates the major characterstisc of the protocols.

  SSF (Scalable Simulation Framework) [22, 23] is designed to model very large-scale networks, which are described in generic modelling language called Domain Modelling language (DML). SSF can be run in a distributed environment (called DaSSE) on a tightly coupled shared-memory symmetric multiprocessor. DaSSF achieves good parallel performance by using a periodic time-stepped approach. All processors can safely process messages between synchronisation cycles without fear of erroneous results due to unsafe events. SSF supports multicast in-channel (many to one communication) as well as multicast out-channel (one to many) and bus-style channel mapping (many to many). SSFNET is the first collection of SF-based models for simulating Internet protocol and networks. The SSFNET libraries include component models for network elements (hosts, routers, network interface card, local area networks) and network protocols (currently IP, UDP, TCP, BGP and OSPF). SSFNET models are self-configuring, that is each SSFNET class instance can autonomously configure itself. SSF has been demonstrated on networks of several hundred thousand nodes. Contributor: Renesys Corporation. Supports: TCP, UDP, BGP protocols. Languages: Java, C++. Outputs: packet-by-packet simulation. Advantages: 1) Open souce code; 2) Free for education; 3) Platform independentent; 4) This has open source code and based on OOP, it can be enhanced in future. Limitation: Dynamic simulation cannot be visualised using this tool.

  TeD (Telecommunication Description Language) [24, 25, 26] is a language for describing telecommunications networks, coupled with an optimistic network simulation engine, based on Georgia Tech Time-warp. The TeD language specification is split into two distinct parts-MetaTeD and "External Language". MetaTeD defines a set of concepts for modelling the dynamic interactions of entities and their compositions. When MetaTeD is appropriately combined with any regular general-purpose programming language (say C++) then complete language is formed. TeD is process-oriented where the number of processes easily exceeds one million, warranting efficient support for large-scale process orientation. It has demonstrated good performance and scalability when modelling ATM cell switches and private network-network interface (PNNI), Internet and wireless network. TeD has been demonstrated on network models consisting of tens of thousands of nodes. Conservative and Optimistic synchronous processes have been implemented in TeD. Contributor: Kalyan Perumalla, Andrew Ogielski, Richard Fujimoto at Georgia Tech. Year: 1996. Supports: TCP/IP, ATM Private Network to Network Interface (PNNI) signaling neworks, Multicasting protocols, wireless networks. Languages: C++ and Meta-language (MetaTed). Outputs: Simulation of models. Advantages: 1) TeD itself is independent of the underlying parallel simulator and can be used with other parallel simulator (e.g. Nops, recently developed at Darmouth); 2) TeD achieves high parallel performance on multiprocessor machine, speeding up the simulation by a factor proportional to N for N-processor machine. Limitation: 1) Limited protocols are implemented; 2) Not freely available.

  USSF (Ultra-Large Simulation Framework) [27, 28] simulator is based on the WARPED simulation engine. WARPED is a parallel discrete event optimistic simulator based on Time Warp. The syntax and semantic of the input topology models for USSF are described in a Topology Specification Language (TSL). Using TSL, large topologies can be built from smaller sub-topologies and sub-sub-topologies. The topology is parsed into an Intermediate Format (TSL-IF). The analysed TSL-IF by the help of static analyser and code generator modules is then used to generate an optimal simulatable network topology. USSF is used to model large networks using a network of Dual-CPU Pentium processors connected by an Ethernet network. Decoupling of the data and states are provided for swapping data and states in and out of the main memory based on demand, freeing up the memory for better performance of the simulation. USSF has been demonstrated on network models of hundreds of thousands of nodes. The current implementation of USSF is in C++language in UNIX system. Contributor: Dhananj Madhava Rao, Philip. A. Wilsey. Supports: LAN, MAN, Languages: C++. Advantages: Very large-scale simulator (hundreds of thousands of nodes

Table 2: Summary of the small and large scale network simulation tools
Tools Usages Target Network Model Outputs Stored Data Format Platform
Ns-2 Educational, Research WAN Simulation of several protocols Text file (ns/tcl format) Unix
Network Workbench Educational, research LAN/MAN Simulation, summary of interlayer statistics Test file Windows, Unix
NetSim Educational LAN (Ethernet) Simulation, histogram of packet queuing delay, offered load, overall throughput (%), actual data rate, average queuing delay, variance of queuing delay, packet transmission rate, and total number of packets Text file Windows
MaRS Educational WAN, link-state and distance vector routing, SPF, ExBF, Segal routing Simulation of application traffic-ftp, telnet, simple, various meters (binary, bar graph, histogram, line graphs, etc), periodic and event updated statistics, throughput, delay, jitter, dropped packets, routing load etc., no trace file, Hope count, utilisation, delay and hope-normalised delay Not known-require further investigation Windows, Unix
OPNET Commercial LAN, Satellite, radio modelling Simulation of several protocols, delay, utilisation etc. performance analysis Text file (C/C++ format) X window
COMNET III Commercial LAN, MAN, WAN Node utilisations, application delays, link delays and utilisation, message delays, packet delays, calls blocked, disconnected and pre-empted, session setup delays etc. Graphical input/output. No text file for modelling Unix
REAL Commercial LAN Statistics such as the number of packets sent by each source of data, number of packet received by sink, the queuing delay at each queuing point, and the number of dropped and retransmitted packets Text file Unix
ATM-TN Specialised LAN, MAN, WAN Simulation of ATM network Text file in Unix system Unix
GloMoSim Specialised Wireless network Simulation of wireless network Text file (two configuration files) Windows, Unix

  C. Specialised Tools

  ATM-TN (ATM Traffic and Network simulator) [29] is designed to characterise cell level network behaviour. The simulator incorporates three classes of ATM traffic source models: an aggregate Ethernet model, an MPEG model, World Wide Web transaction model and six classes of ATMswitch architectures including output buffered, shared memory buffered and cross bar switch models. The ATM-TN simulator can be used to characterise arbitrary ATM networks with dynamic multimedia traffic loads. Call set up and tear down via ATM signalling is implemented in addition to the various types of cell traffic streams generated by voice, video and data. The simulator is built on a simple, efficient simulation language called SimKit, which is capable of supporting both fast sequential and parallel execution. Parallel execution is supported using WarpKit, an optimistically synchronised kernel that is aimed at shared memory multiprocessor platforms. The main design principles of the ATM-TN are: 1) accurately mimic ATM network behaviour at the cell level for specific traffic loads; 2) create a modular extensible architecture; and 3) achieve responsible execution times for ATM networks that consist of hundreds of traffic sources. The structure of ATM-TN consists of the components: traffic models, switch models, an ATM Modelling Framework (MF), SimKit, WrapKit, OSS and the Telecom Modelling Framework (TMF). Several distinct components are to construct traffic models: input traffic sources (e.g. FTP, Telnet, Mosaic, JEPG, MPEG), ATM Adaptation Layer (AAL) for converting source data to cell packets, access control mechanisms (e.g. leaky bucket). There are three types of traffic models are used: MPEG, Ethernet and WWW. The basic switch models in ATM-TN simulator have two components: the control module and the switch fabric. Besides dynamic behaviour of ATM network, analytical outputs can be estimated which include: propagation delay, link capacity in bit/sec, error rate, work load and several parameters. Contributor: Telesim project led by Brian Unger at University of Calgary. Supports: ATM, Data sources (FTP, Telnet, Mosaic, TCP/IP), Video sources (JPEG, MPEG, video-conferencing), WWW model, Qnet model. Languages: C++. Outputs: Link utilisation, QoS, delay analysis. Advantages: 1) Overhead is low; 2) High performance and realistic model; 3) The amount of computation associated with processing such events is low; 4) ATM-TN has a GUI that provides easy simulation scenario configuration, data set organisation, and control over simulation execution and report generation. Limitation: At present, workload is balanced manually.

  Glomosim (Global Mobile System simulator) [30, 31, 32] is a scalable simulation environment for wireless mobile network systems. It is designed using the parallel discrete-event simulation capability provided by PARSEC. The protocol stack includes: models for the channel, radio, MAC, network, transport, and higher layers. It supports TCP, IEEE 802.11 CSMA/CA, MAC, UDP, HTTP, FTP, CBR, Fishleye, LARScheme-1, ODMRP, WRP, DSR, MACA, Telnet, AODV, etc. protocols. There is a visualisation tool VT designed to view and help debugging the protocols. Glomosim currently supports protocols for only wireless network. In the future, it is planned to add functionality to simulate a wired as well as a hybrid network. There are eight files related to a network model designed in Glomosim network simulator. Six input files are used to execute a network model: Configuration file, Nodes file, Mobility file (movement of the nodes and traffic), Routers file, Application file, Berbpsk file (Bit Error Rate). Two files are related to the output of the network model. All of the statistics (selected simulation events) is compiled together into a file called glomo.stat that is produced at the end of the simulation. Trace file is produced (if 'write trace' option is chosen) which can be played as many times as user wants and this is faster than real time. Basically, Glomo.stat and trace files store same set of data. But trace file can be simulated any time, which is not possible with Glomo.stat. Supports: TCP family, UDP, CBR, FTP, HTTP protocols, wireless. Languages: Java, C. Outputs: Dynamic output in VT tool, trace file. Advantages: 1) Easy configurable and fast simulation; 2) Many protocol already implemented; 3) Well documented; 4) Dynamic behaviour can be visualise using VT tool; 5) Large scale simulator; 6) Open source code; 7) can be extended for future. Limitation: 1) Currently wired network is not imnplemented;2) Incomplete API. Table 2 and Table 3 show the summary of small and large scale simulators, and very large scale simulators respectly

Table 3: Summary of the very large-scale network simulation tools
Tool Usages Description Simulation resolutions Scale of operation Stored Data Format Platform
Pdns Educational, Network research Extension to the ns simulator. Works in conjunction with Georgia’s RTIKIT to process events in a correct timestamp(LBTS) order in the case of distributed simulation. The topology is modelled using (GT-ITM) Packet level Hundred of thousands Text file(tcl) Unix
TeD Commercial, Network research Tele-communication networks, TeD (telecommunication Descriptive Language) is a tool that brings automated parallelisation of network simulation by transforming its models into functionally equivalent GTW (Georgia Time- Warp). C++ classes in ns become entities in TeD but ns packet class is implemented as an event. On the negative side, it acknowledges the cost of ‘state-saving’ as the most serious among Time Warp overheads Packet level Tens of thousand TED language (text) Sun Solaris
SSF Commercial, Network research Large-scale networks described in domain modelling language (DML), SSF is a discrete event modelling API designed for very large networks and can execute a million or more concurrent TCP/IP flows. SSFNET models are self-configuring and configuring data is hierarchical structured. SSF architecture has just five generic primary classes IP packet Few hundred thousand DML(domain modelling language) configuration file Linux, Sun Solaris, Unix, Windows
USSF Commercial, research To simulate complex model with over 1 million components, USSF is a framework that runs as an application on an underlying parallel kernel and utilising its services. The kernels include the WARPED based on optimistic PDES and NOTIME (an asynchronous PDES kernel). RTEL was developed to reduce the static size of the application modules and in turn its static memory requirements. Its an ideal candidate for simulating large applications that contain LPs of common description LP (message passing) Few hundred thousand TSL (topology specification language) format(text) Linux


V. Conclusions

  In this paper, a survey of the research work on network MS tools (analytical and simulation only) with a classification is presented. We have tried to review all the recent research work very briefly (because of space limitations) that have been brought to our attention, identified which are the main contributions that have been made, and what are the issues that seem to be open to further research.

  Our overall conclusion from this survey is that parallel, very large-scale network simulators are relatively new and still rapidly evolving field where Internet is trying to map. If all of the network MS tools are integrated and used together then the complementary features can be applied in a network model and justified it [33]. Existing research has certainly made significant contribution in this field. However rapid advances in simulation technology enable to execute highly interactive and dynamic, parallel simulations/animations on the network technology.

References

  1. Bragg A.W. “Which network Design Tool Is Right for You?” IT Professional, September/October, 2000, p23-31.
  2. Cahn R.S. “Wide Area Network Design: Concepts and Tools for Optimization”. Morgan Kaufman Publishers, San Francisco, 1989.
  3. Ma H. (Bob). “Network Design Tool. Computer Science and Engineering”, Washington University in St. Louis, MO., 1989. (http://www.cs.wustl.edu/~javagrp/ network-design- tool.html).
  4. Bachmann D.W., M.E. Segal, M.D. Srinivasan, T. J. Teorey. “NetMod: A Design Tool for Large–Scale Heterogeneous Campus Network”. Center for Information Technology Integration (CITI), University of Michigan. (IEEE J. on Selected Areas in Communications. January 1991, Vol 9, No 1, 1990, p15-24)
  5. Deboo K. “XNetMod: A Design Tool for Large-Scale Networks”. CITI Technical Report TR-93-6. University of Michigan, 1993. (http://www.citi.umich.edu/techreports/reports/citi-tr-93-6.pdf).
  6. Shahbaz M. “Network Design Tool: TND-Tool”. ComNets Annul Report. Communication Networks, Aachen University of Technology, Germany, 1996. (http://www.comnets.rwth-aachen.de/report96/node129. html).
  7. Post. S.D. “Network Planning with a Performance-Prediction Tool”. International Journal of Network Management, Vol 9, Issue 3, 1999, p167-173.
  8. Choi S. “Resource Configuration and Network Design in extensible Networks”. PhD Thesis. Department of Computer Science and Engineering, Washington University, 2003. (http://www.arl.wustl.edu/~syc1/web/paper/ thesis.pdf)
  9. Bresla L., D. Estrin, K. Fall, S. Floyd, J. Heidemann, A. Helmy, P. Huang, S. McCanne, K. Varadhan, Ya Xu and H. Yu. “Advances in Network Simulation”. The VINT Project. IEEE Computer, N5, 2000, p59-67.
  10. Fall K. and K. Varadhan. 2003. The ns Manual. The VINT Project, 13 December 2003. (http://www.isi.edu/nsnam/ns/doc/ns_doc.pdf)
  11. Pullen J.M. “The Network Workbench: Network Simulation Software for Academic Investigation of Internet Concepts”. Computer Networks. Elsevier Science Ltd, March 2000, Vol 3, Issue 3, 2000, p365-378.
  12. Barnett III B.L. “An Ethernet Simulator for Undergraduate Networking”. ACM SIGCSE Bulletin, Vol 25, Issue 1, 1993, p145-150. (Proceedings of the 24th SIGCSE Technical Symposium on Computer Science Education. Indianapolis, Indiana, USA.)
  13. Griffin D. “Traffic Engineering for Quality of Services in the Internet, at Large Scale”. Technical Report, CEC Deliverable Number: 201/UCL/bl, TEQUILA Consortium, 2000.
  14. Alaettinoglu C., Dussa-Zieger, I. Matta, A.U. Shankar and O. Gudmundsson. “Introducing MaRS, a Routing Testbed”, ACM SIGCOMM, pp 95-96.
  15. Riley G.F. and M.H. Ammar. “Simulation Large Networks: How Big is Big Enough?”. Proceedings of 1ST International Conference on Grand Challenge for Modeling and Simulation, San Antonio, TX, 2002, p39-45.
  16. Chang X. 1999. “Network Simulations with OPNET”. Proceedings of the 1999 Winter Simulation Conference, Vol 1, 1999, p307-314.
  17. OPNET Technologies, Inc. 2004. “OPNET Modeller”. (http://www.opnet.com/ products/modeler/home.html)
  18. Ahjua S.P. “Comnet III: A Network Simulation Laboratory Environment For A Course In Communications Networks”. Frontiers in Education Conference (IEEE/ASEE conference), Tempe, AZ, 1998. (http://fie.engrng.pitt.edu/fie98/papers/1205.pdf.)
  19. CACI Products Company. “COMNET III User’s Manual”. January 1995.
  20. Goble J.G. and R. Mills. COMNET III: Object-Oriented Network Performance Prediction. Proceedings of the 1994 Winter Simulation Conference, 1994, p443-445.
  21. Keshav S. “REAL 5.0 User Manual”. Cornell University, Ithaca NY. August 1997. (http://www.cs.cornell.edu/skeshav/real/user.html)
  22. Cowie J.H., D.M. Nicol and A.T. Ogieliski. “Modeling the Global Internte”. Computing in Science and Engineering. Vol 1, Issue 1, 1999, p42-50.
  23. Cowie J.H., D.M. Nicol and A.T. Ogieliski. “Toward Realistic Million–Node Internet Simulations”. International Conference on Parallel and Distributed Processing Techniques and Applications (PDPTA’99), Las Vegas, Nevada, 1999, p2129-2135.
  24. Bhatt S. and B. R. Fujimoto. “Parallel Simulation Techniques for Large-Scale Networks”. IEEE Communication Magazine, 1998, N8, 1998, p42-49.
  25. Perumalla K., R. Fujimoto and A. Ogielski. “TeD - A Language for Modeling Telecommunication Networks”. ACM SIGMETRICS Performance Evaluation Review, Vol 25, Issue 4, 1998, p4–11.
  26. Perumalla K.S. and R.M. Fujimoto. “Efficient Large-scale Process-oriented Parallel Simulations”. Proceedings of the 30th Winter Simulation Conference, 1998, p459-466.
  27. Wilsey P.A. and D.M. Rao. “Simulation of Ultra-Large Communication Network”. Proceedings of 7th International Symposium on Modeling. Analysis and Simulation of Computer and Telecommunication Systems, MASCOT’99, 1999, p112-119.
  28. Wilsey P. A. and D. M. Rao. “An Ultra-Scale Simulation Framework”. Journal of Parallel and Distributed Computing, January 2000, Vol 10, No 1, 2000, p18-38.
  29. Unger B. W., P. Gburzynski and C. Williamson. “A High ATM Traffic and Network Simulator”. Proceeding of the 1995 Winter Simulation Conference, Ottawa, Ontario, 1995, p996-1003.
  30. Cavin D., Y. Sasson and A. Schiper. “On the Accuracy of MANET Simulators”. POMC’02, Toulouse, France, 2002, p38-43.
  31. University of California, Los Angeles. 2001. “GloMoSim Manual (ver. 1.2)”. 07 February 2001. (http://pcl.cs.ucla.edu/projects/glomosim/ GloMoSimManual.html)
  32. Zeng X., R. Bagrodia and M. Gerla. “GloMoSim: A Library for Parallel Simulation of Large-scale Wireless Network”. Proceedings of the 12th workshop on Parallel and Distributed Simulation, Banff, Alberta, Canada, 1998, p154-161.
  33. Rahman M.A., A. Pakstas and F.Z. Wang. “An Approach to Integration of Network Design and Simulation tools”, Proc. of the 8th International IEEE Conference on Telecommunications (ConTEL 2005), Zagreb, Croatia, June 15-17, 2005, p173-180.