Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

5GC$^2$ache: Improving 5G UPF Performance via Cache Optimization

Haonan Jia, Meng Wang, Biyi Li, Yirui Liu, Junchen Guo, Pengyu Zhang

TL;DR

This work addresses LLC-induced performance limitations in cloudified 5G UPFs, where three cache-related problems—leaky DMA, hot/cold mbuf, and cache contention—can severely degrade throughput. It presents 5GC$^2$ache, a cache-optimization framework with three modules (Status Profiler, UPF Configurator, LLC Allocator) that offline-tunes RX descriptors and RX buffer while online isolating DDIO and core caches via a four-state controller and Intel CAT. The approach yields significant gains on a commercial 5G core, achieving up to 76.41 Gbps throughput (a 39.41% improvement over the default) and outperforming ResQ by 29.55%, all without modifying UPF source code. This cache-centric optimization demonstrates a practical path to boost 5G UPF performance in real deployments and will be released as open-source.

Abstract

Last Level Cache (LLC) is a precious and critical resource that impacts the performance of applications running on top of CPUs. In this paper, we reveal the significant impact of LLC on the performance of the 5G user plane function (UPF) when running a cloudified 5G core on general-purposed servers. With extensive measurements showing that the throughput can degrade by over 50\% when the precious LLC resource of UPF is not properly allocated, we identify three categories of performance degradation caused by incorrect LLC usage: DMA leakage problem, hot/cold mbuf problem and cache contention. To address these problems, we introduce the design and implementation of 5GC$^2$ache that monitors the LLC status as well as the throughput performance and dynamically adjusts key parameters of the LLC resource allocation. Our experiments show that 5GC$^2$ache enables a commercial 5G core to increase its throughput to 76.41Gbps, 39.41\% higher than the original performance and 29.55\% higher than the state-of-the-art.

5GC$^2$ache: Improving 5G UPF Performance via Cache Optimization

TL;DR

This work addresses LLC-induced performance limitations in cloudified 5G UPFs, where three cache-related problems—leaky DMA, hot/cold mbuf, and cache contention—can severely degrade throughput. It presents 5GCache, a cache-optimization framework with three modules (Status Profiler, UPF Configurator, LLC Allocator) that offline-tunes RX descriptors and RX buffer while online isolating DDIO and core caches via a four-state controller and Intel CAT. The approach yields significant gains on a commercial 5G core, achieving up to 76.41 Gbps throughput (a 39.41% improvement over the default) and outperforming ResQ by 29.55%, all without modifying UPF source code. This cache-centric optimization demonstrates a practical path to boost 5G UPF performance in real deployments and will be released as open-source.

Abstract

Last Level Cache (LLC) is a precious and critical resource that impacts the performance of applications running on top of CPUs. In this paper, we reveal the significant impact of LLC on the performance of the 5G user plane function (UPF) when running a cloudified 5G core on general-purposed servers. With extensive measurements showing that the throughput can degrade by over 50\% when the precious LLC resource of UPF is not properly allocated, we identify three categories of performance degradation caused by incorrect LLC usage: DMA leakage problem, hot/cold mbuf problem and cache contention. To address these problems, we introduce the design and implementation of 5GCache that monitors the LLC status as well as the throughput performance and dynamically adjusts key parameters of the LLC resource allocation. Our experiments show that 5GCache enables a commercial 5G core to increase its throughput to 76.41Gbps, 39.41\% higher than the original performance and 29.55\% higher than the state-of-the-art.
Paper Structure (23 sections, 10 equations, 19 figures, 4 tables)

This paper contains 23 sections, 10 equations, 19 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Background
  3. 5G Background
  4. DDIO, RX buffer and descriptor
  5. Cache usage in 5G core UPF
  6. Motivation
  7. Leaky DMA problem
  8. Hot/cold mbuf problem
  9. Cache contention problem
  10. 5-G-C-$^2$-a-c-h-e Overview
  11. 5-G-C-$^2$-a-c-h-e Design
  12. UPF Configurator
  13. LLC Allocator
  14. Implementation
  15. Evaluation
  16. ...and 8 more sections

Figures (19)

  • Figure 1: 5G architecture.
  • Figure 2: Each RX descriptor points to a memory buffer (mbuf) in the RX buffer. When receiving a packet, DDIO can use the descriptor to directly load the packet into LLC.
  • Figure 3: Packet reception, loading, processing, and transmission in the 5G UPF.
  • Figure 4: NIC writes packets directly into LLC and then evicts previous packets into DRAM, resulting in DMA leakage.
  • Figure 5: The mbuf becomes cold since not been accessed for a while, resulting in hot/cold mbuf issues.
  • ...and 14 more figures