Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

MemorAI: Energy-Efficient Last-Level Cache Memory Optimization for Virtualized RANs

Ethan Sanchez Hidalgo, J. Xavier Salvat Lozano, Jose A. Ayala-Romero, Andres Garcia-Saavedra, Xi Li, Xavier Costa-Perez

TL;DR

The paper addresses energy inefficiency in virtualized RANs caused by LLC cache contention (noisy neighbors). It proposes MemorAI, an optimization framework that combines per-vBS Digital Twins with an offline-trained neural network classifier to allocate LLC cache ways and minimize total vRAN computing usage. By decomposing the problem per vBS and training offline, MemorAI avoids runtime performance degradation while yielding near-optimal energy savings. Experiments on a PoC vRAN platform show MemorAI outperforming standard allocation strategies, especially under high traffic, demonstrating a scalable path toward energy-efficient 5G virtualization.

Abstract

The virtualization of Radio Access Networks (vRAN) is well on its way to become a reality, driven by its advantages such as flexibility and cost-effectiveness. However, virtualization comes at a high price - virtual Base Stations (vBSs) sharing the same computing platform incur a significant computing overhead due to in extremis consumption of shared cache memory resources. Consequently, vRAN suffers from increased energy consumption, which fuels the already high operational costs in 5G networks. This paper investigates cache memory allocation mechanisms' effectiveness in reducing total energy consumption. Using an experimental vRAN platform, we profile the energy consumption and CPU utilization of vBS as a function of the network state (e.g., traffic demand, modulation scheme). Then, we address the high dimensionality of the problem by decomposing it per vBS, which is possible thanks to the Last-Level Cache (LLC) isolation implemented in our system. Based on this, we train a vBS digital twin, which allows us to train offline a classifier, avoiding the performance degradation of the system during training. Our results show that our approach performs very closely to an offline optimal oracle, outperforming standard approaches used in today's deployments.

MemorAI: Energy-Efficient Last-Level Cache Memory Optimization for Virtualized RANs

TL;DR

The paper addresses energy inefficiency in virtualized RANs caused by LLC cache contention (noisy neighbors). It proposes MemorAI, an optimization framework that combines per-vBS Digital Twins with an offline-trained neural network classifier to allocate LLC cache ways and minimize total vRAN computing usage. By decomposing the problem per vBS and training offline, MemorAI avoids runtime performance degradation while yielding near-optimal energy savings. Experiments on a PoC vRAN platform show MemorAI outperforming standard allocation strategies, especially under high traffic, demonstrating a scalable path toward energy-efficient 5G virtualization.

Abstract

The virtualization of Radio Access Networks (vRAN) is well on its way to become a reality, driven by its advantages such as flexibility and cost-effectiveness. However, virtualization comes at a high price - virtual Base Stations (vBSs) sharing the same computing platform incur a significant computing overhead due to in extremis consumption of shared cache memory resources. Consequently, vRAN suffers from increased energy consumption, which fuels the already high operational costs in 5G networks. This paper investigates cache memory allocation mechanisms' effectiveness in reducing total energy consumption. Using an experimental vRAN platform, we profile the energy consumption and CPU utilization of vBS as a function of the network state (e.g., traffic demand, modulation scheme). Then, we address the high dimensionality of the problem by decomposing it per vBS, which is possible thanks to the Last-Level Cache (LLC) isolation implemented in our system. Based on this, we train a vBS digital twin, which allows us to train offline a classifier, avoiding the performance degradation of the system during training. Our results show that our approach performs very closely to an offline optimal oracle, outperforming standard approaches used in today's deployments.
Paper Structure (17 sections, 1 equation, 13 figures, 1 table)

This paper contains 17 sections, 1 equation, 13 figures, 1 table.

Table of Contents

  1. Introduction
  2. Background: memory in general-purpose computing platforms
  3. Experimental analysis
  4. Cache memory isolation
  5. LLC occupancy and utility
  6. The problem
  7. Problem Formulation
  8. MemorAI
  9. Digital Twin
  10. NN Classifier
  11. Performance Evaluation
  12. Training Evaluation
  13. Digital Twin
  14. NN Classifier
  15. Performance Benchmark
  16. ...and 2 more sections

Figures (13)

  • Figure 1: Aggregated per-core usage with # of vBS instances in our PoC vRAN platform.
  • Figure 2: Energy consumption as a function of the computing load.
  • Figure 3: vRAN testbed.
  • Figure 4: Comparison of the aggregated per-core usage with # of vBS instances showing the "No isolation", the "Pinning" and the "Pinning + LLC isolation" scenarios.
  • Figure 5: Instructions per cycle (IPC) with # of vBSs.
  • ...and 8 more figures