Micro-service deployment method based on intimacy perception and load balancing

Abstract

The invention provides a micro-service deployment method based on intimacy perception and load balancing, comprising the following steps: S1, defining a micro-service deployment problem as a Markov decision process, S2, inputting a directed weighted graph of a micro-service into an AttnSAGE model to obtain a state st, S3, inputting data of the state st into an Actor network to obtain a target sample {st, at, Rt, st + 1}; s4, randomly sampling a plurality of samples in the experience playback pool, and putting the target sample {st, at, Rt, st + 1} into the experience playback pool; and S5, inputting the target sample and the plurality of samples into two Critic networks to calculate a target Q value, and guiding strategy updating by estimating the target Q value under the state st and the action at. According to the method, a micro-service dependency relationship is modeled into a directed weighted graph, and a sorting list comprehensively considering the dependency relationship between services is generated by utilizing community division and calculating the eigenvector centrality of each micro-service.

Images