THESIS
2001
xii, 129 leaves : ill. ; 30 cm
Abstract
The past few years have seen increasing interest in arbitrary topology cell-based local area networks, such as ATM. In these networks, hosts are connected together by an arbitrary graph of communication links and switches. They offer a number of potential advantages relative to other approaches. However to realize these potential advantages, a high performance router or switch is needed to take data arriving on an input link and quickly deliver it to the appropriate output link. In this thesis, we consider switches dealing with fixed-length packets, or cells....[
Read more ]
The past few years have seen increasing interest in arbitrary topology cell-based local area networks, such as ATM. In these networks, hosts are connected together by an arbitrary graph of communication links and switches. They offer a number of potential advantages relative to other approaches. However to realize these potential advantages, a high performance router or switch is needed to take data arriving on an input link and quickly deliver it to the appropriate output link. In this thesis, we consider switches dealing with fixed-length packets, or cells.
We explore the basic architecture of routers and different types of switch fabrics. According to the relative speed of the input ports and the switch fabric, a switch can be classified as input queued (IQ), output queued (OQ) or combined input and output queued (CIOQ). IQ switches have attracted more attention because of low hardware requirement. Virtual Output Queuing (VOQ) is a method to solve the HoL problem of IQ switches and is a practical and high-performance packet switch architecture.
We propose two unicast scheduling algorithms for the VOQ architecture switches and a new architecture and its corresponding scheduling algorithm to handle multicast traffic. We also consider how to integrate them in order to process the real-life traffic composed of both unicast and multicast packets. In order to provide QoS scheduling based on our algorithms, we propose a buffer management method which can be used on input queues to provide fair share of resources among different flows.
Post a Comment