An Architectural Approach for Heterogeneous Data Access in Serverless Platforms

The continuous digitalization and application of modern ICT technologies in the Smart City and Tourism domains are paving the way to new, integrated and connected experiences with strong economic and social impact. However, the integration of services and data coming from different providers is hindered by a rapidly evolving ecosystem of tools and techniques, introducing substantial delays and costs in solution design, deployment, testing, and refinement. Serverless computing is a novel cloud computing model where the customers' business logic, structured as lightweight functions, is automatically put into execution in response to an incoming event. This emergent paradigm promises to be an appealing solution, lowering the development and management barrier for the adaptation, integration, and roll-out of services. However, the ephemeral nature of serverless functions clashes with the necessity of creating and maintaining persistent connections to the various data providers. In this work, we present the Serverless Persistence Support (SPS) service, a novel, distributed, and scalable service implementing an adaptation layer able to improve data access performance from serverless functions. To validate our proposal, we conduct a thorough analysis on a realistic testbed, contrasting various data layer supports and assessing SPS performance when compared to the classic approach where connections and operations are executed inside the business logic. The experimental analysis shows that our solution exhibits better performance both in terms of latency and throughput, while also improving resource utilization.