Experimental research in high capacity communications for data centers

2013年，数据中心每年处理的数据量超过了Zettabyte（或1021字节，或万亿字节）阈值，而社交网络，云计算和大数据应用的巨大进步仍在继续。根据思科全球云指数，数据中心流量的复合年增长率为32%，预计到2020年底将达到14.1泽字节。数据中心内的流量占整个数据中心流量的75%以上。这些趋势正在推动对未来数据中心内每秒PB（Pb/s）速率的研究。Pb/s目标需要多种复用策略和先进的调制，所有这些都在数据中心严格的成本限制范围内。实现Pb/s传输速率需要一个系统，该系统可以同时沿着所有复用途径突破边界：先进的调制格式、频谱高效的操作、多个波长带的覆盖以及模式和/或多个核心中的空间复用。为了保持低成本，解决方案必须与单片集成光学和电气元件的趋势兼容。在实现光学组件的各种技术中，硅光子学特别有吸引力，因为它具有小尺寸、低成本、低功耗和可扩展性的潜力。此外，软件定义网络的出现和对灵活网格体系结构的兴趣限制了可行的Pb/s系统策略。拟议的五年研究项目将侧重于探索如何利用WDM和QAM调制来利用光多路复用的成果，以在短距离（< 5公里）的点对点或网状网络中实现单根光纤上的Pb/s。该项目将把最先进的技术推向几个方向

In 2013, the amount of data processed yearly in data centers surpassed the Zettabyte (or 1021 bytes, or the trillion Gigabyte) threshold, while the massive progress in social networking, cloud computing and big data applications continues. With a compound annual growth rate of 32% per the CISCO global cloud index, data center traffic is expected to reach 14.1 Zettabytes by the end of 2020 . Traffic that remains within data centers accounts for more than 75% of the overall data center traffic. These trends are pushing research for Petabit per second (Pb/s) rates within future data centers. The Pb/s target will require multiple multiplexing strategies and ad-vanced modulation, all within the tight cost constraints of the data center.Achieving Pb/s transmission rates requires a system that pushes boundaries simultaneously along all multiplexing avenues: advanced modulation formats, spectrally efficient operation, coverage of multiple wavelength bands, and spatial multiplexing in modes and/or multiple cores. To keep costs low, solutions must be compatible with trends to monolithically integrated optical and electrical components. Among the various technologies to implement optical com-ponents, silicon photonics is particularly attractive as it holds the potential for small footprint, low cost, low power consumption and scalability . In addition, the advent of software defined networks and interest in flex-grid architectures place constraints on viable Pb/s system strate-gies. The proposed five-year research project will focus on exploring how results in optical multiplexing can be harnessed with WDM and QAM modulation to achieve Pb/s on a single fiber in a point-to-point or mesh network at short distances (< 5 km). The project will push the state-of-the art in several directions