Research on Intelligent Scheduling of Multi-Cloud Serverless Functions and Cross-Platform Interoperability Based on Deep Q-Learning

Abstract

Yufeng Hu and Minmin Liu

Aiming at the challenges faced by serverless computing in multi-cloud environments, this study proposes an intelligent function scheduling algorithm based on deep Q learning (DQN). Kubernetes multi-tenant mechanism is used to simulate multi-cloud environments and achieve seamless interoperability of functions across platforms. The algorithm provides a new solution for serverless computing by optimizing resource utilization, network latency and scheduling costs. Experiments show that compared with traditional static scheduling algorithms, this algorithm has better decision- making in high-dynamic scenarios. Resource utilization is improved by about 33% (such as from 0.6 of traditional static scheduling to more than 0.8 of DQN intelligent scheduling), function response time is reduced by about 40% (such as from 500ms to about 300ms), and call overhead is reduced by about 30% (such as from 100-unit cost to about 70 unit cost). This study provides technical support for the promotion of multi-cloud serverless computing. Although it has limitations, it lays a foundation for subsequent research.