Data Tier Architectures for Enterprise SaaS: A Comparative Performance Study of Tenant Isolation Models

Multi-tenancy is the foundational architectural principle enabling Software-as-a-Service providers to serve multiple enterprise customers efficiently from shared infrastructure. However, the choice of tenant isolation model at the data tier presents critical trade-offs between performance, resource efficiency, and enterprise-grade data protection requirements. This paper presents an empirical evaluation of three primary multi-tenancy data architectures: dedicated database, isolated schema, and shared schema, using a real-world enterprise asset integrity management application deployed in a containerized environment. The case study application, serving clients in the oil and energy sector, was instrumented to measure latency under varying concurrent user loads and data volumes. Results demonstrate that dedicated and isolated schema models deliver superior performance, particularly beyond 200 concurrent users, while shared schema exhibits exponential latency degradation at lower capacity thresholds despite greater resource efficiency. The findings provide actionable guidance for enterprise SaaS architects navigating the tension between operational cost optimization and performance guarantees required by industrial customers. This research contributes empirical data to inform architectural decisions for multi-tenant enterprise applications where regulatory compliance, data sensitivity, and workload characteristics vary across tenant organizations.