Automation of distributed data management in applied microservices package for scientific computations

Abstract

We offer a specialized toolkit for automating both knowledge management when creating an applied microservices package and data accumulating during its application for scientific computations in a hybrid computing environment. The decentralized solving of the declaratively formulated problem is carried out by an active agent group. This group is self-organized by logical inference on the distributed knowledge base of a subject domain. The developed toolkit automates the creating and updating of the local knowledge base of the manager-agent of applied microservices package, as well as the local knowledge bases of distributed computational agents. Local knowledge bases are formed using a description of the interface of computational microservices managed by these agents. Microservice ensembles, corresponding to the active group, are stored in the knowledge base of the manager-agent. The developed toolkit uses this information for testing microservice in the case of its update. In hybrid computing, this toolkit provides synchronizing, archiving, and saving of calculated data. Hybrid infrastructure combines the reliability and availability of using on-premises computers with scaling to the cloud when peak loads occur. The conducted experiments confirmed the effectiveness of the presented approach for solving practically significant scientific problems.