High-performance interconnection networks are among the key factors determining the scalability of supercomputing and intelligent computing systems. Topology serves as the core of scalability-oriented interconnection network design. The design of topology must not only address macro-level requirements from applications and hardware-software systems, but also consider multiple constraints such as router chip port count, number of virtual channels, and packaging density. Significant topological structures from both academia and industry were systematically analyzed and summarized, and a detailed exposition of representative novel topologies was provided. The design challenges of adaptive routing in high-radix networks were examined, the performance and cost of typical topologies were compared, and recommendations for topology selection were discussed. Furthermore, it preliminarily explored future challenges and trends in topology design, including developing cost-effective network topologies tailored to the characteristics of intelligent computing applications, coordinating topology design with building power supply constraints, and integrating and co-designing intra-and inter-supernode network topologies.