A transaction platform for microservices-based big data systems

Abstract

Microservices architecture has increasingly been adopted for building distributed and scalable applications. The premise is that microservices are designed as smaller software components which are easier to be preserved and which enable separation between different components. This paper proposes a new transaction platform for microservices architecture to manage processing of big data stored in a cluster of NoSQL databases. New asynchronous protocols are designed to execute database operations as transactions, and to maintain their correctness and consistency. A prototype system has been developed that simulates London bus service across bus routes. It is evaluated through simulation experiments using big data from ‘Transport for London’ data service in order to analyse effects of transaction processing on response time and throughput in microservices architecture. The transaction platform reliably processes database operations, and enables data availability and consistency in failure-free and failure-prone environments.