Analytical modeling of SCTP and load balancing SCTP

This dissertation focuses on studying and improving congestion control of Stream Control Transmission Protocol (SCTP). The study comes from two approaches: analytical modeling and NS-2 simulation. We developed a steady state throughput analytic model of standard SCTP which includes time-out, secondary path retransmission, slow-start and congestion avoidance phases. Simulations of widely varied scenarios have done on NS-2 and results are compared with predictions from our analytical model. Comparisons show that the analytic model can accurately predict throughput of SCTP for a wide range of packet error rates (PER) and round trip times (RTT). Performance comparison with TCP also have been extensively studied and the results demonstrate that SCTP provides better throughput performance in almost all network environments we have tested especially when PER is high. The results also suggest that SCTP is a better choice of transport layer protocol for error prone network environments such as wireless networks but not necessary for long delay networks such as satellite communications.; Multi-homing is the most significant feature that SCTP provides. But how to leverage the multiple interfaces provided by multi-homed SCTP endpoint for better throughput and reliability is still under intensive investigation. In this dissertation, we propose a mechanism that allows multi-homed SCTP nodes naturally achieve load balancing across multiple paths. Proposed mechanism has been implemented on NS-2. Simulation results show that proposed mechanism can significantly increase the aggregated throughput. A steady state throughput analytic model of the proposed load balancing SCTP has also been developed and the results from analytic model and NS-2 simulation are compared. Comparisons show that our model can precisely predict aggregated throughput of load balancing SCTP as well as standard SCTP.