Research & Papers

WASP framework brings configurable stateful serverless to edge-cloud continuum

Swap WASM runtimes and datastores on the fly without touching code.

Deep Dive

WebAssembly (WASM) has emerged as a lightweight alternative to containers for Function-as-a-Service (FaaS) deployments, but existing WASM-based serverless platforms are tightly coupled to specific execution engines and primarily designed for stateless workloads. This creates a mismatch with the heterogeneous edge-cloud continuum, where hardware resources and application requirements vary widely. Stateful applications—those that maintain data across invocations—are particularly ill-served.

To address this, a team from Politecnico di Milano (Matteo Cenzato, Dario d'Abate, et al.) developed WASP, a configurable framework that brings portable stateful serverless execution to the edge-cloud continuum. WASP abandons monolithic architectures in favor of strictly decoupled, pluggable components: system administrators can swap the WASM runtime (e.g., Wasmtime vs. Wasmer) and the datastore (e.g., in-memory vs. persistent) to fit available resources and application needs—all without altering application code. Configurable lifecycle and caching policies further enable fine-tuning for diverse non-functional requirements like latency sensitivity or memory constraints.

The team's experimental evaluation demonstrates that WASP introduces negligible runtime overhead. By changing runtimes, datastores, and policies, they observed radically different memory and latency profiles, confirming the framework's adaptability to the heterogeneous constraints of the edge-cloud continuum. The work is accepted for presentation at the International Workshop on Scalable Compute Continuum (WSCC 2026) and represents a practical step toward truly portable serverless computing.

Key Points
  • WASP decouples WASM runtime and datastore into pluggable components, enabling swap without code changes.
  • Supports stateful serverless applications with configurable lifecycle and caching policies for diverse requirements.
  • Experimental evaluation shows negligible overhead while exposing radically different memory and latency profiles across runtime/datastore combinations.

Why It Matters

Enables flexible, portable serverless deployments across heterogeneous edge-cloud hardware without vendor lock-in.

📬 Get the top 10 AI stories daily