Serverless Mesh Architectures for Multi-Cloud and Edge

Abstract

Serverless computing is changing the cloud application de- sign by removing the need to design, build, and manage infrastructure, and instead focusing on deploying code that can be elastic and rapid. However, while service meshes have recently been introduced to address the reliability of communication in microservices architectures, the grow- ing adoption of edge computing and multi-cloud strategies require new architectures that can cross different types of platforms. In this paper, we introduce a novel serverless mesh architecture that combines server- less function platforms with a service mesh overlay that spans across clouds and edge sites. We give a brief overview of the serverless and service mesh technologies and review the related work in traditional cloud-centric serverless, edge computing deployments and mesh-based microservice solutions. We then describe a framework where the server- less functions that are located in different cloud/edge environments are connected by a single service mesh that enables dynamic routing, inter- operability, and policy enforcement across the environments. The dia- grams show how the functions are deployed and invoked via the mesh. Simulated evaluation is described for video analytics, IoT sensor data in- gestion, and content delivery use cases, with the potential performance benefits shown in terms of end-to-end latency and throughput. Some of the key challenges of network latency, cold start delays, and data con- sistency issues in distributed serverless computing are identified, along with how the proposed architecture tackles or alleviates them. Last, fu- ture directions are outlined, which include AI/ML-based optimizations for function allocation and tuning, enhanced runtime portability via We- bAssembly (WASM), and open questions for secure, scalable serverless computing from edge to cloud.