作者： Adrian Popescu , R. P. Singh
DOI:
关键词:
摘要: process to open up all three fundamental bottlenecks. For reasons discussed in the dissertation, a direct extension of standard, vertically-oriented OSI model multigigabit optical network environment may not be suitable. Furthermore, design follows new concepts opening electro-optic and service A WDM architecture is used that based on Wavelength-Dedicated-to-Application (WDA) concept. The has an structured multiclass with pipelined for communication. coarse bottleneck. Specific time-synchronous channels, placed different wavelengths, are dedicated applications control mechanisms according their traffic characteristics, e.g., CBR and/or VBR isochronous traffic, computer data transfer (with focus latency), application-specific media access, error handling, flow control, etc. We use separation user-traffic, information from order separate, simplified application-oriented protocols supporting both packetand circuit-switching processing An original solution advanced problem multiple stream (low-level) synchronization WDA network. It includes specific solutions point-to-point transmission, bit (jitter compensation) slot synchronization, as well proper algorithms frame synchronization. Detailed theoretical analysis performance evaluation provided study typical cases, which include one, or multiple, channel(s) channel(s), wavelengths same fiber. Discussion basic implementation issues well. In provide bounded delay jitter requirements loss-free nonisochronous we suggest approach congestion bandwidth allocation, type Rate Controlled Admission Priority Scheduling Service. call this Dynamic Time Sharing (DTS). This guaranteeing parameters (bandwidth requirements) through policer/shaper unit, then optimizing assignment, within network, interest (such jitter, loss). optimization guaranteed by shaper. batch admission policy at edges framing strategy follow (peak bandwidth) characteristics subclasses. On other hand, another (congestion control) (delay/loss) classes. Proper management buffer resources every (switch) node such guarantee diverse interest, regardless statistics, size transmission speed. results clearly indicate feasibility concept candidate future multi-gigabit communication over multiwavelength networks. Such architectural offers choice restricted optics, less electronics processing. reducing latency limits commensurable those given propagation Real-time services, specified requirements, can easily provided. Simple low-processing used, no interference among user-traffic therefore very suitable integration, irrespective conditions. There need technological breakthroughs, wavelength agility, LAN environments developed capable large number supercomputers. disadvantage mainly hardware replication, needed each class.