A Runtime Infrastructure for the Compute Continuum

The vision

For a more comprehensive exposition, consult the following white paper.

The key traits of the envisioned Compute Continuum are the following.

• Interactions, either between applications, components, and users, should leverage Web protocols, such as HTTP/3.
• Edge and Cloud resources should be accessed as a single pool of resources, continuous and seamless.
• Application components should be able to migrate, opportunistically, across the entire resource pool.

Project's roadmap

The construction of a runtime infrastructure for the Compute Continuum will happen in three steps.

• Construction of a node

  • Support live migration in a Wasm runtime:
    A migration-capable Wasm interpreter
  • Three deployment scenarios
    

    (1) Bare-metal, without an OS
    

    (2) On top of a RTOS
    

    (3) Leveraging a general-purpose OS
    

• Construction of a cluster

  • Select and deploy an orchestration engine
  • Extend the orchestrator with Continuum-native functionalities

• Benchmarking

  • Benchmark selection or realization
  • Experimental evaluation over the infrastructure

Scenarios

• High-Performance Computing

  To leverage an increasingly vast pool of computing resources, both in Cloud and Edge, for High-Performance Computing (HPC) applications

• Energy Harvesting Systems

  Migrating computations in the presence of low-energy phases, instead of suspending them, aiming at a higher system utilization

Publications

• E. Tinto, T. Vardanega, A runtime infrastructure for the Continuum of Computing, HPDC ’24 PhD Symposium, PDF.

Contacts