Network simulator design with extended object model and generalized stochastic Petri net.

摘要

A new method is proposed to model and implement a network simulator. To describe complete characteristics of a network system the Extended Object Model (EOM) is used. EOM is an integrated analytical model that completely expresses data structure and control flow of the system. EOM supports a data structure model, which uses the analogy of an object-oriented class and a control model, which uses Generalized Stochastic Petri-Net.; To implement control in objects requires a class unit called EOC++, which dynamically links the object model to the control model. A Virtual Transition Table (VTT) in the EOC++ provides a flexible mechanism, which couples loosely the object structure and behavior.; The Extended Object-Oriented Simulator (EOOS ) provides extended objects with a discrete time-driven simulation platform. This platform employs the Replicating Real Traffic (RTT) method to increase the accuracy of traffic source modeling. This method employs the Reading SNMP Variable tool ( RSV) and Simple Network Management Protocol (SNMP) to measure the statistics of network components.; GSPNs as the control model are checked for executional properties and states reduced to only those that are realizable. The GSPN is checked using reachability trees and markings.; Validation tests were performed on a space-division ATM switch, the Knockout switch, with Constant Bit Rate (CBR) traffic sources using GSPN tools and Extended Object-Oriented Simulator (EOOS). These tests check the switch performance parameters. The results were compared with analytical methods under uniform and nonuniform loads. The results showed an accuracy of 0.008 and 0.37 Root Mean Square Error (RMSE) for GSPN and simulator results respectively.; Tests were performed to evaluate the EOOS accuracy in modeling traffic sources. The tests were performed on a real network (the ATM network system of MU campus backbone) to measure the performance statistics: traffic rate, network utilization, network throughput, buffer occupancy, cell loss ratio (CLR), inter arrival time (IAT), and cell transfer delay (CTD). Tests showed an average of 10% accuracy improvements in generating traffic using Replicate Real Traffic (RTT) technique compared to stochastic distribution technique, which is commonly being used.; The method for producing the simulation of the ATM network shows an approach to provide a more general simulation network system design. The simulation framework, consisting of kernel and EOC++, is completely independent of the extended objects and can be used to simulate any network system. Therefore, more general simulations require the creation of only a new set of extended objects: modeling the Extended Objects (EOs) by GSPN, implementing each EO with EOC++ to create the simulation entities, and linking them to the simulator kernel to generate the executable file.