Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

POSEIDON : Efficient Function Placement at the Edge using Deep Reinforcement Learning

Prakhar Jain, Prakhar Singhal, Divyansh Pandey, Giovanni Quattrocchi, Karthik Vaidhyanathan

TL;DR

POSEIDON addresses the edge function placement problem under resource constraints and dynamic workloads by combining a PPO-based DRL agent for placement with a MILP-based router for traffic routing. It minimizes both network delay $T$ and function running cost $C$ using a configurable trade-off $\alpha$, and learns through a reward that normalizes these objectives while penalizing infeasible placements. Empirically, POSEIDON matches or surpasses state-of-the-art in delay with substantially faster decision times (about $3$–$7$ ms, ~16x faster than baselines) and favorable cost-delay profiles, demonstrating practicality for responsive edge deployments. The approach offers a scalable, adaptive solution for edge function placement that can complement existing routing optimization frameworks in heterogeneous, nomadic MEC environments.

Abstract

Edge computing allows for reduced latency and operational costs compared to centralized cloud systems. In this context, serverless functions are emerging as a lightweight and effective paradigm for managing computational tasks on edge infrastructures. However, the placement of such functions in constrained edge nodes remains an open challenge. On one hand, it is key to minimize network delays and optimize resource consumption; on the other hand, decisions must be made in a timely manner due to the highly dynamic nature of edge environments. In this paper, we propose POSEIDON, a solution based on Deep Reinforcement Learning for the efficient placement of functions at the edge. POSEIDON leverages Proximal Policy Optimization (PPO) to place functions across a distributed network of nodes under highly dynamic workloads. A comprehensive empirical evaluation demonstrates that POSEIDON significantly reduces execution time, network delay, and resource consumption compared to state-of-the-art methods.

POSEIDON : Efficient Function Placement at the Edge using Deep Reinforcement Learning

TL;DR

POSEIDON addresses the edge function placement problem under resource constraints and dynamic workloads by combining a PPO-based DRL agent for placement with a MILP-based router for traffic routing. It minimizes both network delay and function running cost using a configurable trade-off , and learns through a reward that normalizes these objectives while penalizing infeasible placements. Empirically, POSEIDON matches or surpasses state-of-the-art in delay with substantially faster decision times (about – ms, ~16x faster than baselines) and favorable cost-delay profiles, demonstrating practicality for responsive edge deployments. The approach offers a scalable, adaptive solution for edge function placement that can complement existing routing optimization frameworks in heterogeneous, nomadic MEC environments.

Abstract

Edge computing allows for reduced latency and operational costs compared to centralized cloud systems. In this context, serverless functions are emerging as a lightweight and effective paradigm for managing computational tasks on edge infrastructures. However, the placement of such functions in constrained edge nodes remains an open challenge. On one hand, it is key to minimize network delays and optimize resource consumption; on the other hand, decisions must be made in a timely manner due to the highly dynamic nature of edge environments. In this paper, we propose POSEIDON, a solution based on Deep Reinforcement Learning for the efficient placement of functions at the edge. POSEIDON leverages Proximal Policy Optimization (PPO) to place functions across a distributed network of nodes under highly dynamic workloads. A comprehensive empirical evaluation demonstrates that POSEIDON significantly reduces execution time, network delay, and resource consumption compared to state-of-the-art methods.

Paper Structure

This paper contains 15 sections, 12 equations, 4 figures, 3 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Problem and Solution Overview
  3. Solution overview
  4. POSEIDON
  5. Function Placement
  6. Routing policies
  7. State Update and Solution Tuning
  8. Evaluation
  9. Experimental Setup
  10. Delay and cost analysis (RQ1)
  11. Decision time analysis (RQ2)
  12. Impact of solution tuning (RQ3)
  13. Threats to Validity
  14. Related Work
  15. Conclusions

Figures (4)

  • Figure 1: POSEIDON architecture.
  • Figure 2: Effectiveness of the approaches with respect to various metrics for the small payload
  • Figure 3: Effectiveness of the approaches with respect to various metrics for the large payload
  • Figure 4: Cumulative count of invalid placements versus iterations of solution tuning for small payload ($\alpha=0$)