High-speed interconnection network topology discovery method and device, medium and high-performance computing system

Abstract

The invention discloses a high-speed interconnection network topology discovery method and device, a medium and a high-performance computing system. The topology discovery method comprises the following steps: reading opposite-end mark information from a network interface chip of a host node and adding the opposite-end mark information into a seed queue; gradually iteratively searching the seed queue for the opposite-end mark information, recording the opposite-end mark information into an adjacency table, and constructing a source routing table entry pointing to the opposite-end network interface chip; and finally, comparing the actual topological structure expressed by the adjacency list with the expected topological structure to discover a wrong port connection relationship. The invention also discloses a device, corresponding to the high-speed interconnection network topology discovery method, a medium and a high-performance computing system. According to the method, the network topology structure can be found quickly, and the actual topology structure can be compared with the expected topology structure to find out the wrong port connection relationship, so that the debuggingtime of the interconnection network is shortened, and the construction efficiency of the interconnection network is improved.

Description

technical field [0001] The invention mainly relates to the high-speed interconnection network field of high-performance computing systems, and in particular to a high-speed interconnection network topology discovery method, device, medium and high-performance computing system. Background technique [0002] The life cycle of a high-performance computing system is mainly composed of three parts: one is the construction time of the computing system, including the hardware and software installation, configuration, debugging, testing and other processes of each subsystem; the other is the normal operation time of the computing system, that is, the computing system can The time to provide services for high-performance computing applications; the third is the system fault processing time. Some minor system faults can be diagnosed and processed online, while individual serious faults need to be processed after the application is offline or even shut down for maintenance. As the size...

AI Technical Summary

Problems solved by technology

[0006] The technical problem to be solved by the present invention is to provide a high-speed interconnection network topology discovery method for the problem that the interconnection network topology discovery technology based on the InfiniBand protocol of the existing high-performance computing system cannot perform interconnection network topology structure inspection. Device, medium and high-performance computing system, the present invention supports fast realization of network topology discovery, and supports comparison of actual topology structure and expected topology structure to find wrong port connection relationship, thereby shortening interconnection network debugging time and improving interconnection The efficiency of network construction can not only facilitate the realization of topology structure inspection in the network chip hardware logic and network management software, but also meet the needs of network commissioning and testing personnel to quickly check the network topology structure

Images