Optical Burst Switching - A New Paradigm

点击翻译按钮获取中文摘要

URL: http://www.acsu.buffalo.edu/~qiao/nsf_abs.html There are two basic drivers for optical internetworks. One is the explosion of the multimedia (mainly data) traffic over the Internet, especially the World-Wide-Web, which as many discovered recently, can be bursty at all time scales and various multiplexing levels. The other is the continuing advances of WDM optical networking technologies, which offer many opportunities to streamline both software (protocols) and hardware (electronic equipments) for reduced latency and cost. The objective of this research is to explore a novel switching paradigm for WDM (Wavelength Division Multiplexed) optical networks, called Optical Burst Switching (OBS). OBS uses an offset time between a control (or set-up) packet and its corresponding data burst when making one-way delayed reservation, which facilitates all-optical data transfers and distinguishes it from other one-way reservation schemes and label/tag switching schemes. By combining the best of the coarse-grained optical circuit switching (via wavelength routed lightpaths) and the fine-grained optical packet/cell switching, while avoiding their shortcomings, OBS can efficiently support applications requiring many short-lived (or bursty) sessions during which a substantial amount of data needs to be transferred at a high bit rate and with a low end-to-end latency. More importantly, it will help realize the vision of building a flexible, efficient and bandwidth-abundant fiber-optic internetworking infrastructure capable of providing ubiquitous services through TCP/IP (as well as ATM and other existing and future protocols). In this project, we will focus on the design and performance evaluation of efficient OBS protocols that can reduce burst dropping probability, and support flow and congestion control as well as multicasting, priority and fault-tolerant routing. We will also study how upper layer protocols (e.g. IP) interact with OBS, investigate related optical-layer control and managemen t issues such as resource provisioning, protection and restoration, and in addition, compare OBS with other burst switching and lable/tag switching techniques. Finally, we will examine the synchronization problem and the cost-effective designs of the interface between the upper layers and the optical layer as well as the designs of the WDM switches.