DRL-based Slice Placement under Realistic Network Load Conditions

TL;DR

This paper presents a DRL-based network slice placement method that incorporates heuristics for convergence control, optimized for large-scale, dynamic network environments with fluctuating traffic loads.

Contribution

It introduces a heuristic-controlled DRL approach tailored for realistic, large-scale networks with non-stationary traffic, demonstrating improved stability and performance over standard DRL methods.

Findings

Higher and more stable performance compared to non-controlled DRL.

Effective in large-scale, non-stationary network scenarios.

Suitable for online learning with volatile requests.

Abstract

We propose to demonstrate a network slice placement optimization solution based on Deep Reinforcement Learning (DRL), referred to as Heuristically-controlled DRL, which uses a heuristic to control the DRL algorithm convergence. The solution is adapted to realistic networks with large scale and under non-stationary traffic conditions (namely, the network load). We demonstrate the applicability of the proposed solution and its higher and stable performance over a non-controlled DRL-based solution. Demonstration scenarios include full online learning with multiple volatile network slice placement request arrivals.