Dataplane Performance Comparison of 5G gNB Implementations

Overview

This dataset provides a comprehensive collection of network performance metrics captured through a specialized testbed and methodology designed for the research community. The data is structured into three distinct categories: Throughput measurements, One-Way Delay (OWD) measurements, and OWD analysis conducted under heavier traffic loads. To balance human readability with computational efficiency, the dataset is provided in a hybrid format.

Data Structure and Features

The dataset is organized into three primary feature sets:

1. Throughput Metrics: Provided in JSON format (generated via iPerf3), this set includes mbpsactual_uplink/downlink, mbpsoffered_uplink/downlink, meanjitterms_uplink/downlink, and meanloss_uplink/downlink.2. OWD / Latency Metrics: Provided in CSV format parsed from traffic captures, these files contain packet-level telemetry including epoch timestamps, source/destination IP addresses (Inner and Outer), packet size (Bytes), sequence numbers, inter-arrival times (iat), relative time (trel), path distance (pdist), and packet counts (pnpak).3. Metadata and System Logs: This set includes experimental context such as gNodeB (gnb) identifiers, bandwidth (bw), TDD ratio, scenario descriptions, and repetitions. It also features physical layer metrics like Modulation and Coding Schemes (dl_mcs, ul_mcs) and Signal-to-Noise Ratio (ue_snr, gnb_snr).

Contents

The repository includes the processed JSON and CSV datasets, raw system logs detailing CPU usage and SNR measurements, and the original processing scripts used to integrate these metrics for evaluation. This multi-layered approach allows researchers to validate the results or perform independent analysis on the relationship between system load and network performance.