Multi-Cloud Deployment Strategies: Portability, Latency, and Vendor Lock-In Trade-Offs for Stateful Microservices

Abstract

Organisations adopting multi-cloud strategies face fundamental tensions between workload portability, inter-cloud latency, data sovereignty, and operational complexity. This paper analyses four multi-cloud deployment patterns�Active-Active, Active-Passive, Cloud-Burst, and Geo-Partitioned�for stateful microservices across AWS, Azure, and GCP. Using a purpose-built benchmarking harness, we measure P99 latency, failover time, replication lag, and monthly cost for each pattern under synthetic and replay traffic loads. Active-Active achieved lowest failover time (1.8 s) but incurred 2.4� cost premium. Geo-Partitioned offered the best latency-cost balance for data-sovereignty-constrained workloads. Decision matrices and reference architectures are provided for architects.