D3QN-based secure scheduling of microservice workflows in cloud environments

Abstract

In recent years, microservice architecture has been widely applied in software design, including scientific workflow scheduling in the cloud, due to its loose coupling characteristics. With the development of container technology, its fast startup and low overhead make it the preferred deployment technology for microservices. However, a key issue with current microservice architecture is the risk of exposing private and sensitive data. Existing work has researched and explored new architectures for microservices, yet the trade-off between privacy and time has not fully considered still. To fill this gap, we first designed a security model that considers the security levels of users and cloud providers, as well as proposed a workflow scheduling algorithm that satisfies both budget constraints and task privacy requirements. We transform the secure scheduling problem into a Markov decision process (MDP) and propose the Dueling Double DQN-based microservice workflow scheduling Algorithm (D3MWS), which can not only minimize the completion time under budget constraints, but also meet the security requirements of workflow tasks. Finally, well-known workflow applications in cloud computing were selected for experiments. The experimental results showed that the proposed algorithm, while meeting the security requirements of microservices, reduced the average task completion time of each workflow by 10.91% compared to other algorithms under different task quantities.