Packet Management System in Single-Server Queue Networks with Poisson Arrivals Using a Treap-Based Model

Abstract

There are enormous studies on packets’ congestion control in queue networks. However, not enough research work had been done on the management of packets being dropped in queue networks such that these packets could be re-transmitted without accruing unnecessary costs to the system. Since packets' losses amount to a waste of network resources, this study proposes a model with which packets dropped are kept in a tree structure and later re-transmitted to the server once it becomes available. A Treap-Model Congestion Control System (TMCCS) was proposed to prevent network congestion by ensuring that packets arriving in the system when the queue network is saturated are managed to avoid congestion. The model was benchmarked with Random Early Detection with Reconfigurable Maximum Dropping Probability (RRMDP). OMNeT++ was used as a simulation framework while datasets were generated randomly. Simulation results indicated that while the average throughput for RRMDP was 93.8mbs, that of TMCCS was 103.6mbs. Similarly, while the packets’ average queue size for RRMDP was 55.1(x 10−3)mbs, that of TMCCS was 54.3(x10−3)mbs respectively. Consequently, it was concluded that TMCCS is more efficient in the management of packets in queue networks with regard to network throughput and average queue size.