讲座名称：Directed percolation routing: Unlock multi-path diversity for ultra-reliable and low latency services
讲座地点：腾讯会议直播（ID：753 853 670 密码：123456）
Lin Cai分别于2002和2005在加拿大滑铁卢滑铁卢大学获得电气和计算机工程硕士学位和博士学位，现任维多利亚大学电气与计算机工程系教授。她是NSERC E.W.R. Steacie Memorial Fellow， IEEE Fellow，加拿大皇家学会院士，被评为2020“Star in Computer Networking and Communications”。她的研究兴趣涵盖通信和网络的多个领域，重点是支持新兴多媒体通信和物联网的网络协议和架构设计。
In a mesh network, the number of "shortest" paths in terms of hop-count increases exponentially w.r.t. network size. To fully explore path diversity for ultra-reliable and low-latency (URLL) services, we advocate a directed percolation routing (DPR) strategy, i.e., the message will be forwarded toward the destination hop by hop, so long as the next hop neighbor is closer to the destination. Using DPR, what is the probability that a packet can reach its destination is the network connectivity problem. Closely related to directed percolation, network connectivity for mesh networks is very complicated to calculate. The existing state-of-the-art can only calculate connectivity for a lattice network up-to-the size of 10 x 10. In this talk, we introduce a new approach, named Hop-State Algorithm (HSA). HSA is simpler and more scalable to calculate connectivity of lattice network, up-to 16x16 (containing more than $600$-million shortest paths), and can handle general network topology and heterogeneous links. Based on HSA, we can choose a subset of network links to use to improve network efficiency and by-pass congested links, while ensuring reliability. As a case study, we apply DPR to ensure URLL services in low earth orbit (LEO) satellite backbone networks. Using Starlink constellation (with 1,584 satellites) as an example, we show the superior performance using DPR for ensuring high reliability and low latency services.