5G and emerging 6G testbeds generate large volumes of timestamped measurement data that must be analysed to enable reliable and reproducible evaluation of network performance. A key problem is that the original 5GENESIS Analytics module was designed for InfluxDB v1, which causes compatibility issues in InfluxDB v2 environments, particularly regarding authentication and the query interface. In this thesis, the 5GENESIS Analytics module is modernised to operate with InfluxDB v2 while preserving existing REST interfaces and workflows, and an offline LSTM-based approach is evaluated for predicting network performance from KPI time-series data. The problems are addressed by upgrading the Data Handler with token-based authentication, Flux-based data retrieval, and improved metadata handling, and by training LSTM models on exported experiment datasets. The results show stable operation of the analytics stack with InfluxDB v2 and consistent values across the dashboard, the Data Handler API, and Grafana. The prediction workflow shows that LSTM models can capture overall time-series trends and provide meaningful forecasts for metrics such as throughput and latency, although short-term spikes are often smoothed.
Hanan et al. (Thu,) studied this question.