A machine learning approach for performance-oriented decision support in service-oriented architectures

Enterprise IT performance can be improved by providing reactive and predictive monitoring tools that anticipate problem detection. It requires advanced approaches for creating more agile, adaptable, sustainable and intelligent information systems. Service-oriented architecture (SOA) has been used in significant performance-based approaches by information system practitioners. Organizations are interested in performance-based decision support along the layers of SOA to maintain their sustainability for service reuse. Reusability is a very important aspect of Service-based systems (SBS) to analyze service or process reuse. This helps in achieving business agility to meet changing marketplace needs. However currently, there are many challenges pertaining tothe complexities of service reuse evolution along SBS. These challenges are related to the sustainability of service behavior during its lifecycle and the limitations of existing monitoring tools. There is a need for a consolidated classified knowledge-based performance profile, analytical assessment, prediction and recommendation. Therefore, this paper provides a semantic performance-oriented decision support system (SPODSS) for SOA. SPODSS provides recommendations for suggesting service reuse during its evolution. SPODSS is supported by five building blocks. These blocks are data, semantic, traces, machine learning, and decision. SPODSS classify data, validate (analytical assessment, traces, semantic enrichment) at different time intervals and increased consumption and prediction based on consolidated results. It handles the dynamic evolution of SBS and new or changed user requirements by ontology development. Finally, SPODSS generates recommendations for atomic service, composite service, and resourceallocation provisioning. To motivate this approach, we illustrate the implementation of the proposed algorithms for all the five blocks by a business process use case and public data set repositories of shared services. Sustainability and adaptability of service-based systems areensured by handling new business requirements, dynamicity issues and ensuring performance. Performance criterion includes functional suitability, time behavior, resource utilization, and reliability in terms of availability, maturity, and risk.