In modern organisations, an increasing number of internal processes are implemented as automated workflows that coordinate various tasks and services. Traditional engines are usually deployed on virtual machines or physical servers and require continuous operational effort: provisioning capacity, handling failures, performing upgrades and monitoring infrastructure. Public cloud providers offer a rich portfolio of managed services that could, in principle, be composed into a fully managed workflow orchestration platform, but there is limited guidance on how to do this in a systematic, cost-effective way.
This thesis presents the architecture of a serverless workflow management tool built on Amazon Web Services (AWS). Workflows are described declaratively and executed by an event-driven core that enforces dependencies, timeouts, concurrency limits, error handling and human interactions, all while keeping application code isolated from low-level infrastructure concerns.
The solution is evaluated through a cost benchmark against an exemplary on-premises deployment taken from the literature. By mapping the reference system to an equivalent AWS deployment of the proposed architecture, the thesis demonstrates that the serverless approach can provide substantial cost savings together with improved scalability and reduced operational overhead.