小尺度空间内光互联多维复杂网络理论和关键技术

Applying optical networking technologies in large scale computing systems has been an irresistible trend which is driven by the IT industry's sustainable development requirements. Optical interconnections used in large scale computing systems typically covers a distance from several centimeters to about one kilometer, which is much shorter than that in the telecommunication scenarios. Therefore, considering the characteristic of data communication in large scale computing systems, this research project investigates the optical interconnection complex network in small space by utilizing complex network analysis theory. Studies are to be carried out around two scientific problems while the main research topics include optical interconnection based complex network topology construction method, new optical network modeling and performance assessment mechanisms, fast routing and load balancing strategies, as well as application-driven on-demand logical topology provisioning scheme. We focus on exploiting the benefits of multi-dimensional resource scheduling capabilities in the optical domain to construct logical topologies with eminent agility. In particular, we creatively propose the idea of building a deterministic small world network with the so called "multi-dimensional space folding" approach. Based on the proposed network model, we also innovatively depict a corresponding "multi-dimension penetration" routing scheme and a "flow field" based load balancing strategy.

以面向大数据应用的大规模计算系统为典型场景，研究小尺度空间内（数厘米至一千米范围）光互联网络技术的应用，符合科学技术发展和社会经济可持续发展趋势。本课题针对大规模计算系统内数据交换网络的需求特点，以光互联网络技术为基础，以复杂网络理论的应用和拓展为设计主线，研究小尺度空间内光互联复杂网络的构建理论及关键技术。课题围绕两个主要科学问题开展研究，内容包括研究光互联复杂网络的拓扑结构和相应实现方式、建立新的网络模型和性能评价体系、研究路由和流量均衡策略，以及提出面向应用任务的动态拓扑提供机制等。课题着眼于发挥光的多维度资源在组网方面的灵活性，创新性地提出了以"多维空间折叠"思想构建"确定性小世界拓扑"的网络模型，以及在此基础上的"多维平面穿越"路由机制和基于"流量场"的负载均衡策略等。

本项目顺应光互联走向小尺度空间、应用于大规模计算系统的科学发展趋势，瞄准大数据处理能力进一步提升的应用需求，围绕“基于光互联在小尺度空间内构建多维复杂网络的机制和方法”和“光互联多维复杂网络的模型和网络调控机制”这两个关键科学问题，结合光互联技术特征和复杂网络理论，开展了五项理论研究：1）面向大规模计算系统的光互联复杂网络逻辑拓扑结构；2）多维复杂网络拓扑基于光互联技术的实现方式；3）光互联多维复杂网络的模型和动态性能评价方法；4）跨维度的路由算法和基于复杂网络参数分布规律的流量均衡策略；5）与上层应用紧耦合的光互联复杂网络逻辑拓扑构建策略。项目取得的成果如下：1）结合光的多维交换资源优势，创新性提出基于空间、时间、频率三种维度组合构建小世界复杂网络拓扑，以适应大规模计算平台的发展需求；2）提出基于多维拓扑辅助图和维度权重相结合的光互联多维复杂网络信息描述模型；3）针对复杂网络中的路由算法和流量均衡问题，提出全局混合复杂网络路由算法和流量均衡策略；4）提出点对间和区域间按需拓扑提供机制；5）搭建了多维光互联复杂网络仿真平台和小规模的多维光互联复杂网络实验平台。最终，项目通过开展系列仿真分析和实验，完善并验证了小尺度空间内构建大规模光互联复杂网络的理论和方法。项目成果有助于推进大规模计算系统的可持续发展和广泛应用。