Performance Evaluation of Burst Assembly Modeling for Optical Burst Switched Network

Abstract

Optical Burst Switching (OBS) is considered as a promising switching technique for the next generation of optical networks. In optical burst switching, the wavelength resources are shared between different connections. At the edge of an OBS network, the packets coming from legacy networks (e.g., IP, ATM networks) are aggregated according to their destinations and form large optical data bursts. The burst are further transmitted and switched in the OBS network. Optical Burst Switch (OBS) allows variable burst formation time and burst sizes so as to optimally utilize the network resources. In this thesis using existing Engset and delay model I have studied the effect of mean packet arrival time, round trip time, fiber capacity, number of buffers etc. on the burst length and burst blocking probability based on simulations. To reduce the burst loss ratio partial burst scheme is proposed. The traffic calculation is based on average traffic rate considering fiber capacity, number of wavelength, number of class and burst length instead of static traffic. The results of partial burst and traffic model are compared with existing Engset model. The proposed model gives lower burst blocking ratio as compared to existing model for the same amount of waiting time. For example at 60 ms of waiting time we obtained a burst loss ratio of -10, -12, -19 for Engset model, Partial burst model and Traffic model respectively. Further the stream line effect has been studied. The simulation result for the same shows that though it has some higher loss ratio but it gives benefit of the concatena- tion of burst which are having same destination at core node. For 90 ms of waiting time burst length is 3356 kb. Burst length is independent of probability scheme. For same value of burst length we get the burst loss ratio of -12 and -13 for the streamline model and existing model respectively. By reducing burst loss ratio and analyzing results of model we can choose burst length and burst aggregation time according to traffic such that minimum resources are required in network. In this way a concept that guaranteed transmission without immediate reservation and explicit release can be realized.