Serverless Computing Framework for Enterprise ERP Background Job Processing in Cloud SAP Environments

Abstract

Enterprise ERP systems execute vast volumes of background work—batch processing, data synchronisation, report generation, interface workloads, and scheduled tasks—that traditionally run on continuously provisioned infrastructure sized for peak demand. This paper presents and evaluates a serverless computing framework for executing and managing ERP background jobs in cloud-based SAP environments. The proposed architecture triggers compute resources dynamically in response to workload demand, enabling event-driven execution, automatic scaling, and consumption-based resource utilisation, thereby reducing dependence on always-on infrastructure. It targets the principal ERP background workload classes while incorporating workload prioritisation, fault tolerance, monitoring, and secure integration to improve scalability, cost efficiency, and operational resilience for large-scale processing. Using a design-science methodology triangulating production-representative telemetry, a controlled before/after comparison against an always-on baseline, and a cost analysis, the study quantifies compute-cost efficiency, scaling latency, job success and fault-tolerance behaviour, and prioritisation fairness under mixed workloads. Results indicate that event-driven, consumption-based execution reduced idle compute cost substantially, scaled elastically to bursty batch demand, and preserved job reliability through retry and idempotency mechanisms, while workload prioritisation protected latency-sensitive jobs under contention. The study also characterises cold-start latency and its mitigation, a central trade-off of serverless execution. Contributions include a serverless reference architecture for ERP background processing, an empirical cost–performance characterisation, and transferable lessons on prioritisation, fault tolerance, and cold-start management. Limitations include single-organisation, observational measurement and workload-class scoping, which constrain generalisation. The work advances practice by showing that consumption-based, event-driven execution can replace over-provisioned batch infrastructure for enterprise ERP while meeting reliability and governance expectations.