A Scalable Communication Integrated Optical Network (SCION) for Bandwidth Greedy Multimedia Applications

该提案为可扩展的集成光网络引入了一种原始的体系结构方法，该网络由具有不同拓扑结构的“LAN”段组成，通过桥连接。研究人员首先研究了两种最有前途的拓扑结构，环形和折叠巴士，它们由一座桥直接连接。由于每种情况下的同步需求和实现问题各不相同，因此在互连网络的不同区域中可能会首选某些拓扑结构。研究人员还建议考虑在几十个子网之间建立一个骨干网作为互连结构。在每个子网（环或总线）中，数据包分别插入和从光纤介质中取出，而在骨干槽中，路由用于在子网对之间交换信息。网桥架构是本提案的主要焦点之一，它将研究涵盖所有主要选项的网桥——无光缓冲的网桥，带光缓冲的网桥（SDL-bridge），以及用于控制网桥操作和争用解决的相关协议，包括插槽覆盖解决方案。桥接网络节点的节点结构必须考虑：收发器的类型、网络的光接口和发送器的缓冲组织。发射器和接收器可以是可调的或固定的，两个设备中至少有一个可调，以提供节点之间的完全连接。这些节点架构选项将与新协议一起进行研究，以支持处理桥接操作所需的不同“排序算法”，以避免产生电子瓶颈。槽路由相关的协议和控制是影响全局网络性能的关键因素。为了在本地流量和网络流量之间分配插槽，需要一个插槽管理协议。将评估两种可能性：一种是固定插槽分配协议，根据预期流量和网络条件以预设顺序分配插槽；另一种是动态插槽分配协议，根据每个网段的瞬时提供负载分配插槽。分析模型和仿真工具将广泛用于评估不同解决方案的性能。

This proposal introduced an original architectural approach for a scalable integrated optical network consisting of "LAN" segments with varying topologies interconnected by bridges. The investigators first investigate the two most promising topologies, rings and folded buses, interconnected directly by a bridge. Since in each case synchronization requirements and implementation issues vary, certain topologies may be preferred in different areas of the interconnected network. The investigators also propose to consider a network backbone that serves as an interconnect structure among dozens of subnetworks. In each subnetwork (rings or buses) packets are individually inserted into and extracted from the fiber medium, while in the backbone slot routing is used to exchange information between pairs of subnetworks. The bridge architecture is one of the main foci of this proposal which will be investigated covering all main options - a bridge without optical buffering, a bridge with optical buffering (SDL-bridge), and associated protocols for controlling the bridge operation and contention resolution, including the slot-override solution. The node architecture of bridged network nodes must take into account: the type of transceiver, the optical interface to the network, and the buffer organization at the transmitter. The transmitters and the receivers can be tunable or fixed, with at least one of the two devices being tunable to provide full connectivity among the nodes. These node architecture options will be investigated together with the new protocols to support different "sorting algorithms" needed to deal with bridge operation to avoid the creation of electronic bottlenecks. The slot routing related protocols and control are a key factor in the global network performance. In order to allocate slots between local and internetwork traffic a slot management protocol is necessary. Two possibilities will be evaluated: a fixed slot allocation protocol, which allocates slots in a preset order, determined by expected traffic and by network conditions, and a dynamic slot allocation protocol, which allocates slots according to the instantaneous offered load of each network segment. Analytical models and simulation tools will be extensively used to assess the performance of the different solutions.