CONGESTION CONTROL IN ASYNCHRONOUS TRANSFER MODE NETWORKS

Abstract

Asynchronous Transfer Mode (ATM) is the target switching technique for the future public Broadband Integrated Services Digital Networks (B-ISDN). The switch fabric provides the essential muting and buffering functions, but ATM switching systems also require management and control functions necessary for the efficient operation of the network. Congestion is defined as the state of network elements in which network is not able to meet the negotiated performance objective for the established connection. ATM connections can be multiplexed deterministically or statistically. Deterministic multiplexing is done by allocating resources in accordance to peak rate. Here congestion is eliminated and delay is also very small. While this approach is safe but average utilization of bandwidth is poor. In statistically multiplexing, the peak rate of all connections is above the link capacity. In this case if the large number of connections are very bursty, then all of them may be assigned to the same link in the hope that statistically they will not burst at the same time. If some of them do burst simultaneously, there is sufficient elasticity so that the burst can be buffered up at the input. This partly offsets the probability that at one time, offered input at the switch may become higher than the link capacity. However, a longer duration of congestion might result in cell loss and higher delay. ATM networks have an inherent risk of network congestion if they are based on statistical multiplexing. To take care of this problem various congestion control schemes are reported in the literature. The success of ATM networks depends on the development of effective congestion control schemes. These schemes are responsible for maintaining an acceptable QoS level that is delivered by the network. In managing the wide range of traffic types and performance requirements, the key challenge is the equitable and efficient allocation of network resources. Since real-time VBR is inherently bursty, due to unpredictable fluctuations congestion can occur frequently. Therefore an ii important issue in the congestion control of ATM networks is how to handle the conditions of a large number of cells being in transit between two ATM switching nodes. In this thesis work efforts have been made to develop congestion control schemes in ATM networks. Various schemes, which are outcome of the work have been proposed / studied under the followings: i. Congestion control for VBR service in non-blocking ATM switches ii. Congestion control in ATM network using fuzzy approach iii Congestion control for multicast bursty traffic in shared memory ATM switch iv Congestion control in wireless ATM network The performance analysis of schemes includes various Quality of Service (QoS) parameters such as throughput, average cell delay, and cell loss probability. L Congestion control for VBR service in non-blocking ATM switches: In non-blocking ATM switches, when cells are served according to First-In-First=Out (FIFO) strategy, then due to Head-Of-Line (HOL) blocking the performance of the switch is degraded. HOL blocking limits the throughput of each input port to a maximum of 58.6% under uniform random traffic and much lower than that for bursty traffic. Traffic streams in the real world are often characterized as bursty. Most of the Application level Data Units (ADUs), such as video frame, are too large to be encapsulated into a single 53-byte ATM cell and must be segmented into a sequence of cells in order to be transmitted over ATM networks. As a result, consecutive arriving cells in a burst are strongly correlated by having the same destination which addresses the same output of the switch. Keeping this point in mind we have simulated and analyzed performance of non-blocking multiple input ATM switches for VBR bursty traffic using Parallel Iterative Matching (PIM) technique as described in [5]. In this technique, each input port maintains a separate queue for each output port, during a single time slot a maximum of one cell per input port can be transferred, and maximum ofone cell per out put port iii can be received. The switch operation is based on the PIM algorithm to find the maximal matching. Maximal matching is used to determine which inputs transmit cells over the switch to which outputs in the current time slot. It has been observed by simulation that the technique increases throughput and reduces not only mean cell delay but cell loss probability also. These results are very much suitable for providing better QoS for real time VBR bursty traffic applications. ii. Congestion control in ATM network using fuzzy approach: Congestion is a result of a mismatch between the network resources (buffer space, processing and transmission capacity) and the amount of traffic admitted for transmission. Consequently, congestion prevention can be interpreted as the problem of matching the admitted traffic to the network resources. This, in turn, could be viewed as a classical problem of feedback control i.e. matching the output to the input of dynamical systems. Fuzzy logic system have been successfully applied to deal with congestion control related problems in ATM networks and have provided a robust mathematical frame work for dealing with real world imprecision. A fuzzy logic based scheme has been proposed for dynamic feed-back threshold. In this scheme, out-put buffer is divided into various equal number of parts(N) viz. two, three, and four for the purpose, then the feedback had applied after 50%, 33%, and 25% of the buffer space i.e. when N=2, 3, and 4 respectively. Depending upon which threshold has been crossed, the network gets a mild warning, a stem warning or an ultimatum. A gradual change is more intuitive here, this has been incorporated with fuzzy logic. In applied fuzzy scheme, burst length as well as buffer occupancy are represented by triangular functions. After fuzzification and defuzzification process, the percentage blocking to offered at that particular buffer occupancy level and at given burst length has been determined. The performance of the proposed scheme has been compared with that of the constant threshold and dynamic feedback threshold schemes. It has been shown that the fuzzy scheme improves the performance with reference to major parameters of QoS namely - throughput, average delay and cell loss probability. iv iii. Congestion control for multicast bursty traffic in shared memory ATM switch: Shared-memory ATM switches play a leading role in practical, experimental implementation of ATM. Multicast capability is going to play a key role in the design of ATM broadband switch. The management of multicast traffic in shared memory ATM switches is of particular interest. Congestion control is an important phenomena, because ATM is connection-oriented and support real-time service. The concept of threshold helps in maintaining a fine balance between the number of cells propagated and the average delay observed by a typical cell in a switch. We have studied and simulated the effect of various threshold schemes to control congestion for multicast bursty traffic in shared memory ATM switch, and proposed a fuzzy scheme for the purpose. In proposed Fuzzy Threshold Scheme, wherein line occupancy as well as buffer occupancy are represented by fuzzy sets, and represented by triangular functions. After fuzzification and defuzzification process, the percentage allocation space to offered at that particular line occupancy and at given buffer occupancy can be determined. The simulation results obtained for evaluation the performance of various congestion control schemes under unicast (90%) and multicast (10%) mixed bursty traffic load. The results have showed that there is a trade-offs exits between cell delay and cell loss rate parameters. The proposed scheme gives minimum cell delay (unicast and multicast) at the cost of higher cell loss rate. iv. Congestion control in wireless ATM network: In a Personal Communication Networks (PCN), the covered geographical area is typically partitioned into a set of microcells. Each microcell has a Base Station (BS) to exchange radio signals with wireless mobile terminals. Due to the limited range of wireless transceivers, mobile users can communicate only with BSs that reside within the same microcell at any instance. The number of handoffs/handovers during a call will increase as the cell radii decrease, thus affecting the QoS. Frequent handoff in wireless/mobile networks introduces a new paradigm in the area of network congestion and admission control. The increase in processing load due to demand for service and fast handoffs to mitigate the propagation effect, a high speed backbone network for the PCN to connect BS is required. The ATM technology, which has recently emerged to be a predominant switching technology, is suited to be an infrastructure to interconnect the BSs of the PCN. To support network-wide handoff, new and handoff call requests will compete for connection resources in both the mobile and backbone networks. Handoff calls require a higher congestion related performance, i.e., blocking probability, relative to new calls because forced terminations of ongoing calls due to hand-off call blocking are generally more objectionable then new call blocking from the subscriber's perspective. A hybrid scheme has been proposed for handover in ATM-based PCN, which combines queuing and reservation schemes. In the scheme, FIFO and Measurement Based Priority Scheme (MBPS) queuing discipline [158] and Reserved Channel Scheme (RCS) [79] is used. This scheme gives handovers higher priority than queuing or reservation schemes. When reservation is applied on both radio and backbone channels, it leads to significant improvement in QoS. After applying proposed scheme there is a remarkable reduction in Forced Termination Probability (FTP) at the cost oftolerable Call Blocking Probability (CBP).