MODELING AND SIMULATION OF INTERCONNECTED WIRELESS/WIRED ATM NETWORKS

Abstract

In this thesis, flow control issues in interconnected wireless/wired ATM networks are studied. Various simulation models are established which form a testbed to evaluate the performances of various Explicit Rate (ER) flow control techniques. The intelligent congestion control scheme (ICCS) and dynamic intelligent congestion control scheme (DICCS) is implemented and evaluated in the wireline ATM networks. A 'BS(base station)-to-BS' flow control architecture with the DICCS algorithm in it is proposed to provide ABR flow control in interconnected wireless and wired ATM networks.

The ATM Forum has voted to adopt a closed loop rate-based mechanism for congestion control of the Available Bit Rate (ABR) service[8]. DICCS is an enhancement to the ICCS proposed by Siu and Tzeng[21]. The key idea of this technique is for each congested switch to estimate the 'optimal' cell rate for each VC bottlenecked at the switch with a small number of computations and without using state information on a per-VC basis. Two queue length thresholds CQT and DQT are used to indicate normal and extreme congestion status respectively. When extreme congestion occurs, the cell rate at which the source is allowed to transmit data is reduced abruptly. The modification to this scheme in the DICCS is: Upon extreme congestion status, instead of reducing the cell rate of a VC by a fixed proportion, the cell rate is reduced according to the actual queue length. The longer the queue, the more quickly the source is asked to reduce its cell rate. This can be advantageous for DICCS in case of congestion due to a, sudden and short increase in the cross traffic bandwidth. On the other hand, DICCS may need a longer time to deal with the persistent congestion because the reduction of rate is not as abrupt as that in ICCS. Compared to ICCS, DICCS achieves better fairness, higher throughput and better link utilization, but results in a larger queue length. The performance of DICCS is dependent on the threshold size and several other parameters. DICCS keeps the simple and low computational cost merits of ICCS. No floating point computation is required in this scheme.

In the interconnected wireless and wired ATM network, the existence of the low data rate, error prone radio media poses a challenge to the flow control schemes. The feedback information for flow control may be received incorrectly or even lost after being transmitted on the radio link. This will affect the effectiveness of any flow control scheme. To solve these problems, the "BS-to-BS" flow control architecture is proposed in this thesis. In this architecture, the resource management cell is generated by the base station at an ABR connection's source end, turned back by the base station at the VC's destination end, and finally received by the base station at the source end. The feedback RM cell is then used to estimate the ACR value for the VC and the proper number of transmission slots is allocated to the source of this VC according to the estimated ACR. Since no flow control information is transmitted on the wireless link, the above mentioned problems are avoided. The DICCS is integrated into this BS-to-BS architecture. The simulation studies of this BS-to-BS flow control scheme in several network models show that the scheme utilizes the radio resource efficiently, avoids cell loss on buffer overflow at the BS and achieves fairness among mobile terminals in a wireless segment and among ABR applications to one mobile destination.