Cross-data-center network task scheduling method based on reinforcement learning

Abstract

The invention discloses a cross-data-center network task scheduling method based on reinforcement learning. The method mainly solves the problem that task self-adaptive real-time scheduling and three-resource balanced and effective use are realized in a cross-data-center network in a reinforcement learning mode. The method comprises the following specific steps: 1, generating a training data set;2, generating a state space and an action space of reinforcement learning; 3, calculating a reward value of a feasible action in the action space; 4, training a reinforcement learning task schedulingmodel; And 5, scheduling the tasks arrived in the cross-data-center network in real time. Task scheduling across the data center network is achieved through the reinforcement learning method, three-dimensional resource balance and effective utilization serve as the target, the performance of the cross-data center network is optimized, and the resource utilization rate of the cross-data center network is increased.

Description

technical field [0001] The invention belongs to the technical field of communication, and further relates to a network task scheduling method across data centers based on reinforcement learning in the technical field of wired communication networks. The present invention can be applied to a cross-data center network composed of multiple data centers to realize the scheduling of user tasks, so as to satisfy users' requests for computing resources, memory resources, and hard disk storage resources in the cross-data center network when completing tasks. Realize efficient allocation of network resources across data centers. Background technique [0002] With the emergence of technologies such as 5G, Internet of Things, machine learning, and AR / VR, large-scale deployment of cross-data center networks, sudden increase in network traffic, real-time changes in network status, and diversified business requirements have created new challenges for cross-data center networks. Task sche...

AI Technical Summary

Problems solved by technology

The disadvantage of this method is that it does not realize the effective utilization of multi-dimensional resources by quantifying the degree of fragmentation of multi-dimensional resources, resulting in resource fragmentation and affecting the work efficiency of the data center; this method belongs to a fixed heuristic strategy and lacks network The interaction of the real-time state of the environment cannot adaptively optimize the scheduling strategy in the case of dynamic changes in the network environment and task requirements, and is limited

Images