Microservice Remodularisation of Monolithic Enterprise Systems for Embedding in Industrial IoT Networks

Abstract

This paper addresses the challenge of decoupling "back-office" enterprise system functions in order to integrate them with the Industrial Internet-of-Things (IIoT). IIoT is a widely anticipated strategy, combining IoT technologies managing physical object movements, interactions and contexts, with business contexts. However, enterprise systems, supporting these contexts, are notoriously large and monolithic, and coordinate centralised business processes through software components dedicated to managing business objects (BOs). Such objects and their associated operations are difficult to manually decouple because of the asynchronous and user-driven nature of the business processes and complex BO dependencies, such as many-to-many and aggregation relationships. Here we present a software remodularisation technique for enterprise systems, to support the discovery of fine-grained microservices, which can be extracted and embedded to run on IIoT network nodes. It combines the semantic knowledge of enterprise systems, i.e., the BO structure, with syntactic knowledge of the code, i.e., various dependencies at the level of classes and methods. Using extracted feature sets based on both semantic and syntactic dependencies, K-Means clustering and optimisation is then used to recommend microservices, i.e., redistributions of BO operations through microservices from BO-centric components of enterprise systems. The approach is validated using the Dolibarr open source ERP system, in which we identify processes comprising both "edge" operations and requestresponse calls to the Cloud-based enterprise system. Through experimentation using Amazon GreenGrass deployments, simulating IIoT nodes, we show that the recommended microservices demonstrate key non-functional characteristics, of high execution efficiency, scalability and availability.

Evolution of Microservices Identification in Monolith Decomposition: A Systematic Review

“…Identifying microservices involves considering factors such as service boundaries, service quality, and the absence of exhaustive microservices listings [52], [89]. Propositions: Researchers can leverage existing research, such as microservices migration approaches [47], [50], [68], [82], [88], [94], high-quality code summary generation [59], misuse identification [92], and the use of relational topic models for examples [87], as foundations for enhancing microservices identification tools (EC-11). Modern migration techniques should explore hybrid approaches (EC-12) that combine API-side learning with client-side learning [78] and domain adaptation techniques (EC-13) to address out-of-vocabulary problems, a current challenge in microservices identification [103].…”

“…With the emergence of machine learning approaches as potential solutions to key microservices identification issues [8], [21], [47], [58], questions arise regarding the suitability of current approaches for user adoption, their applicability to all issues, and the need for performance enhancements before widespread tool adoption (UC-3). Fuzzy and ambiguous intent (UC-4) and the rapid evolution of software services using microservices, such as IoT devices, present challenges in the evolution of microservices [17], [82]. Propositions: Effective microservices engineering should aim to resolve technical issues stemming from microservices and bridge the knowledge gap between microservices developers and users (UC-5).…”