Fault-tolerant routing method for large-scale computer system

In order to solve the technical problem that in a traditional fault-tolerant routing method, when an optical module breaks down, the duration of an interconnection fault in a system is too long, the invention provides a fault-tolerant routing method for a large-scale computer system, and the method comprises the following steps: 1, recording a fault switch as a source switch i, enabling i to be aserial number of the switch, enabling i to be greater than or equal to 0, carrying out the fault-tolerant routing based on an uplink, and enabling the corresponding flow of the source switch i to be uniformly distributed to other available links, so as to achieve the routing fault tolerance; 2, carrying out fault-tolerant routing based on the downlink, and avoiding a fault port by modifying the routing of the uplink of the source switch i; step 3, ending. The fault-tolerant routing is reconstructed based on the uplink and the downlink respectively, network resources do not need to be increased, and the number of hops