This work presents a scalable and high performance prediction protocol for optical networks. In the proposed protocol, we develop a mathematical model to maintain the stability of a network system by prediction based on the traffic temporal locality property. All the critical factors, including transceiver tuning time, propagation delay, and processing time for dealing with control packets, are considered in the proposed prediction protocol. Furthermore, our protocol can resolve the bottlenecks attributed to control signaling and electronics processing. The performance evaluation reveals that the proposed scheme can yield the higher bandwidth efficiency and incur a lower packet delay than those of the TDM and conventional reservation schemes. Also, the proposed protocol can flexibly support any scaled network system such as MANs or LANs.

Non-real-time (NRT) services such as nrt-VBR, ABR and UBR traffic are intended for data applications. Although NRT services do not have stringent QoS requirements for cell transfer delay and cell delay variation, ATM networks should provide NRT services while considering other criteria to ensure an excellent performance such as cell loss ratio (CLR), buffer size requirement and service fairness. Service fairness means that networks should treat all connections fairly. That is, connections with low arrival rates should not be discriminated against. In addition, given a fixed buffer size for a connection, reducing the maximum number of cells in a buffer during the lifetime of a connection can lead to a low CLR due to buffer overflow. Thus, these criteria should be considered as much as possible when designing a cell scheduler to provide NRT services. Whereas most of the conventional cell scheduling schemes are usually appropriate for one performance criterion, but inappropriate for another one. In this work, we present a novel cell scheduling scheme, called buffer minimized and service fairness (BMSF), to schedule NRT services in ATM networks. Using probability constraints and selecting a connection with the longest buffer size to transmit first allow BMSF to attain a satisfactory performance with respect to maximum buffer size requirement, CLR, and service fairness in terms of the maximum buffer size and cell waiting delay criteria. Simulation results demonstrate that BMSF performs better than some conventional schemes in terms of these criteria, particularly when NRT services have diverse arrival rates. Thus, the BMSF scheme proposed herein can feasibly schedule NRT services in ATM networks.

In this paper a high-performance copy network named PPCN is proposed for large scale ATM switching systems. The proposed copy network consists of multiple planes of the P2I Copy Networks(PCN) arranged in parallel. The PCN planes are designed based on the P2I multistage interconnection networks (MINs). A single PCN plane is itself a preliminary self-routing copy network which, however, is not a non-blocking one. A novel dispatcher is designed to dispatch input cells to the PCN planes such that no internal blocking nor output contention arises during the cell replication procedure and the offered load can be shared in an efficient way. The architecture of the PPCN provides flexibility for the maximum fanout for an input cells. In a PPCN system, the maximum fanout for an input cells is determined only by the number of interconnection stages within the PCN planes, independent of the input size of the system. The performance of the PPCN is studied under uniform traffic. It is shown that a small constant number of PCN planes are sufficient for a PPCN system to achieve an acceptable low overflow probability regardless of the system size. The hardware complexity of an N N PPCN is O(N log2 K) and the length of the routing tag is O(log2 K) bits, where K is the maximum fanout for an input cell. The storage complexity of the translation tables adopted in an N-inlet PPCN is O(N), which is much lower than that of the previously proposed ones.

Real-time services, including constant bit rate (CBR) and real-time variable bit rate traffic (rt-VBR), have become increasingly important owing to the rapid proliferation of multimedia applications. A cell multiplexing method capable of handling real-time traffic should satisfy related quality of service (QoS) requirements, including cell transfer delay (CTD), cell delay variation (CDV) and cell loss ratio (CLR). In this paper, we present an efficient cell multiplexing method, called longest delay beyond expectation (LDBE), to schedule real-time and non-real-time traffic in ATM networks. For the real-time traffic, LDBE scheme can minimize the CDV, and reduce the CLR and CTD, particularly when different CDV tolerance (CDVT) values are applied at each node along the path of a connection. Simulation results demonstrate that the proposed LDBE performs better than other multiplexing methods regarding these CLR, CDV and CTD criteria for real-time traffic. Furthermore, the proposed LDBE is also suitable for scheduling non-real-time traffic by providing a low CLR for non-real-time variable bit rate (nrt-VBR) and minimizing the CTD for unspecified bit rate (UBR) traffic.