Object Oriented Hybrid Network Simulation

Thumbnail Image
TR_94-41.pdf(3.04 MB)
No. of downloads: 412
Publication or External Link
Baras, John S.
Atallah, George C.
Karne, Ramesh K.
Murad, A.
Jang, Kap D.
As the complexity and diversity of networks have grown, simulation has proved an important tool in their design, analysis, testing and performance estimation of networks. Hybrid networks involve a variety of network elements - both mobile (e.g., satellites, mobile radio) and fixed nodes (e.g., switches, hubs, network gateways) linked via varied broadcast, multicast and point-to-point communication channels. Because of their complex nature, design and evaluation of hybrid networks is a particularly complicated task. Major requirements of a hybrid network simulation tool are (i) Flexibility and Adaptability - to accommodate all kinds of hybrid networks and protocols, (ii) Advanced Network Visualization Techniques - to clearly visualize complex communication network systems, (iii) Data Management - to organize and analyze the vast quantities of data generated in a typical simulation run, and (iv) Distributed Implementation - to fully utilize available computing resources to speed up simulation. This paper describes the design and functional description of an Object-Oriented Hybrid Network Simulation tool. Its object oriented design and implementation (in C++) allows flexibility through incorporation of new, user specified network elements, protocols and functional blocks. Advanced visualization techniques are combined with the graphical user interface to allow better visualization of complex network structures. A sophisticated geographical database is also incorporated to aid terrestrial mobile, and satellite network systems. To handle and effectively analyze the vast quantities of data generated, an object-oriented database is incorporated into the simulation. In addition to network simulation, the tool is also designed to serve other needs. An interface is provided to allow the user to run real network applications over the simulated network, allowing network application designers to judge the performance of their applications over various network configurations. Incorporation of a database allows computation of network performance dynamically. A network management tool receiving network performance data both from the actual and the simulated network may use the simulation data to make a long term prediction of the actual network behavior to perform long-term network management.