SCOPE: Stateful Serverless in the Edge-to-Cloud Continuum

Abstract Distributed applications increasingly demand reduced response times to improve user experience. The edge computing paradigm responds to this need by exploiting the resources at the edge of the network to handle user requests close to where they originate, thus avoiding the cost of communication with a remote cloud infrastructure. Most state-of-the-art edge computing solutions build on the serverless model, where developers implement their application as a composition of functions, and the runtime manages and abstracts away deployment and execution in the edge-to-cloud continuum. However, current approaches focus predominantly on stateless functions, which limits their applicability to applications that require robust state management. Our research aims to overcome the above limitations by proposing SCOPE, a stateful FaaS programming model and execution platform for the edge-to-cloud continuum. The SCOPE programming model introduces the concept of sessions to represent portions of the application state that are relevant to a given user or group of users. Serverless functions can create, access, and modify sessions associated to the users who invoked them. The SCOPE platform automatically and dynamically decides where to store sessions in the edge-to-cloud continuum and where to execute functions, balancing between proximity to clients -- to reduce response time -- and availability of resources -- not to overload edge devices with limited resources. Our evaluation shows the benefits of using SCOPE in heterogeneous edge-to-cloud deployments in terms of response time as well as the ability of SCOPE to adapt to mutating operational conditions.