Resource allocation and QoS guarantee in high-speed networks

Resource allocation is a key issue in Broadband Integrated Services Digital Networks (B-ISDN). It provides a tool to guarantee the quality-of-service (QoS) requirements of connections while at the same time it allows efficient utilization of network resources. Resource allocation operates on at least two time scales: call level and packet level. At the call level, when a new connection request arrives, a set of connection admission control conditions are tested at each switch along the path of the connection. The new connection is accepted only if there are enough resources to satisfy the requirements of both the new connection and all existing connections. At the packet level, based on the performance requirements of the connections, the packet service discipline at each switch selects which packet is to be transmitted next. Usually, different service disciplines result in different admission control algorithms. A complete solution is needed to specify both the service discipline and the associated call admission control conditions. In this thesis, new algorithms for resources allocation at both time scales are developed to achieve better performance. At the call level, in order to maximize the number of connections that can be accepted, an optimization algorithm that minimizes the total required bandwidth of the system is developed. At the packet level, we propose a new packet service discipline named Credit-Based Fair Queueing (CBFQ) which can be implemented at the switching nodes of the B-ISDN. In CBFQ, we assign a simple counter to each queue so as to store the credit information. According to the counter values and some other simple information, we can decide which queue should be served next. CBFQ can overcome the implementation complexity in the sorted priority queue structure of the scheduling algorithms proposed by other researchers