Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Service Orchestration in the Computing Continuum: Structural Challenges and Vision

Boris Sedlak, Víctor Casamayor Pujol, Ildefons Magrans de Abril, Praveen Kumar Donta, Adel N. Toosi, Schahram Dustdar

TL;DR

It is shown how Active Inference, a concept from neuroscience, can support self-organizing services in continuously interpreting their environment to optimize service quality and outline a research roadmap toward resilient and scalable service orchestration in the Computing Continuum.

Abstract

The Computing Continuum (CC) integrates different layers of processing infrastructure, from Edge to Cloud, to optimize service quality through ubiquitous and reliable computation. Compared to central architectures, however, heterogeneous and dynamic infrastructure increases the complexity for service orchestration. To guide research, this article first summarizes structural problems of the CC, and then, envisions an ideal solution for autonomous service orchestration across the CC. As one instantiation, we show how Active Inference, a concept from neuroscience, can support self-organizing services in continuously interpreting their environment to optimize service quality. Still, we conclude that no existing solution achieves our vision, but that research on service orchestration faces several structural challenges. Most notably: provide standardized simulation and evaluation environments for comparing the performance of orchestration mechanisms. Together, the challenges outline a research roadmap toward resilient and scalable service orchestration in the CC.

Service Orchestration in the Computing Continuum: Structural Challenges and Vision

TL;DR

It is shown how Active Inference, a concept from neuroscience, can support self-organizing services in continuously interpreting their environment to optimize service quality and outline a research roadmap toward resilient and scalable service orchestration in the Computing Continuum.

Abstract

The Computing Continuum (CC) integrates different layers of processing infrastructure, from Edge to Cloud, to optimize service quality through ubiquitous and reliable computation. Compared to central architectures, however, heterogeneous and dynamic infrastructure increases the complexity for service orchestration. To guide research, this article first summarizes structural problems of the CC, and then, envisions an ideal solution for autonomous service orchestration across the CC. As one instantiation, we show how Active Inference, a concept from neuroscience, can support self-organizing services in continuously interpreting their environment to optimize service quality. Still, we conclude that no existing solution achieves our vision, but that research on service orchestration faces several structural challenges. Most notably: provide standardized simulation and evaluation environments for comparing the performance of orchestration mechanisms. Together, the challenges outline a research roadmap toward resilient and scalable service orchestration in the CC.
Paper Structure (26 sections, 1 equation, 4 figures)

This paper contains 26 sections, 1 equation, 4 figures.

Table of Contents

  1. INTRODUCTION
  2. STRUCTURAL DEFINITION
  3. Problem Hypotheses
  4. Abstract Solution Design
  5. PRELIMINARY IMPLEMENTATION
  6. BEHAVIORAL MARKOV BLANKETS
  7. ACTIVE INFERENCE CYCLE
  8. COLLABORATIVE ORCHESTRATION
  9. PRELIMINARY RESULTS
  10. SUITABILITY OF THIS IMPLEMENTATION
  11. RESEARCH CHALLENGES
  12. SYSTEM & ACTION SIMULATION
  13. Baseline Standardization
  14. Digital Twinning
  15. MODEL UPDATE & INFERENCE
  16. ...and 11 more sections

Figures (4)

  • Figure 1: Three-level implementation for autonomous service orchestration in the Computing Continuum: we interpret current service states (e.g., why was an SLO violated) by training behavioral Markov blankets (MBs) from processing metrics; agents continuously optimize the service operation according to their internal understanding of the environment (i.e., their MB) and current SLOs. Lastly, to optimize SLO fulfillment throughout the CC, we compose MBs to quantify dependencies between services and hosting devices.
  • Figure 2: Missing an ecosystem for pretraining agents in lower-fidelity environments (e.g., simulation) implementing common baselines and infrastructure; agents are deployed in physical devices and pretraining environments improve through real-world feedback.
  • Figure 3: Updates to orchestration policies (e.g., from $\pi_1$ to $\pi_2$) must balance the impact to performance, and tolerate inconsistent model rollout; agents must minimize their execution overhead according to the context, and keep the user in the loop to develop new metrics and align with user intents.
  • Figure 4: Client wants to rent and combine the nearest infrastructure from multiple vendors; agents must orchestrate and cooperate towards global objective, despite limited visibility; ideally, this allows seamless interplay with client-owned devices.