Dataset for: Trade-offs Between GDPR Compliance and User Experience in Cross-Border Cloud AI Inference Services

Dataset: GDPR Compliance and Multi-Region AI Inference Performance Overview This repository contains anonymised performance datasets collected for the study: “Trade-offs Between GDPR Compliance and User Experience in Cross-Border Cloud AI Inference Services.” The data supports an empirical evaluation of how regulatory-driven data localisation (GDPR-compliant vs non-compliant regions) influences latency, throughput, and user-perceived performance in cloud-based AI inference systems deployed on Amazon Web Services (AWS). The datasets were used to generate the quantitative results, figures, and dashboards reported in Sections 4 and 5 of the accompanying manuscript. Dataset Contents The dataset is organised into two primary components: 1. Synthetic Load Testing Data (`k6/`)This directory contains results from automated synthetic benchmarking conducted using the k6 load testing framework. Structure: k6/├── source=eu-west-2/└── source=us-east-1/ Each subdirectory contains multiple JSON files representing individual benchmark runs. Each file includes metrics such as:- Request latency (mean, p90, p95)- Throughput (requests per second)- Success and error rates- Source region and target region- Workload type (self-hosted or managed inference)- Request payload classification (light or heavy) These datasets were used to analyse region-local versus cross-region inference behaviour under controlled workloads. 2. Real User Monitoring Data (`rum/`)This directory contains anonymised real user monitoring (RUM) events collected during interactive testing of the inference system. Structure: rum/├── sourceCountry=GB/└── sourceCountry=US/ Within each directory, JSON files are further organised by the region that served the inference request. Each RUM record includes:- User-observed end-to-end latency- Source country (UK or US)- Served cloud region- Inference workload type- Performance rating classification- GDPR compliance status These datasets were used to assess user-perceived performance and validate whether synthetic benchmark trends align with real-world interaction patterns. 3. Derived Visualisations (`visualisations/`) These are dashboards derived from the datasets using AWS Quicksights which with datasource connected to the s3 bucket where datasets resides. visualisations/ ├── GDPR Performance Overview Dashboard.pdf ├── Performance Trends Analysis Dashboard.pdf └── Regional Comparison Analysis Dashboard.pdf These dashboards are derived visual summaries and are not required to reproduce the experiments. Data Collection and Processing All infrastructure was deployed using identical configurations across regions. Synthetic and real-user data collection periods were aligned temporally. All datasets were exported in structured JSON format directly from automated pipelines. No aggregation or post-processing was applied prior to archival, ensuring traceability from raw data to reported results. Privacy and Ethics All datasets are fully anonymised. No personal data, identifiers, IP addresses, prompts, or user-generated content are included. The study was conducted using synthetic workloads and controlled user sessions. The data complies with GDPR principles and institutional research ethics guidelines. Reproducibility The datasets are sufficient to reproduce:- Latency and throughput comparisons across regions- Managed vs self-hosted inference performance trends- Cross-region routing overhead analysis- User-perceived performance patterns Infrastructure code and application source code are not included but are available from the corresponding author upon reasonable request. License This dataset is released under the **Creative Commons Attribution 4.0 International (CC BY 4.0)** license. Citation If you use this dataset, please cite the associated article and the Zenodo DOI provided with this record. Contact For questions regarding the dataset or methodology, please contact the corresponding author via the details provided in the associated publication.