[DSC DACH 24] How Big Data and AI are transforming Telecommunications and empowering the Connected World - Thomas Hodi

This talk explores the transformative role of Big Data and AI in shaping the future of telecommunications, with a focus on real-world examples that highlight their impact on everyday lives. As the Product Manager for SARA, A1 Telekom Austria Group’s cutting-edge network analytics solution, I will share insights into how our innovations enable smart energy savings in Slovenia, in-home connectivity in Brazil or safe unmanned drone operations in Austria.

1. How Big Data and AI are transforming Telecommunications and empowering the Connected World DSC DACH 24 | Thomas Hodi, A1 Telekom Austria Group

2. Yushu, 2017 DSC DACH 24 | Thomas Hodi

3. The Future of Mobile Communications (according to GenAI ) Prompt: photorealistic picture of a 5G-powered remote surgery Prompt: 5G connected cow photo-realistic resolution DSC DACH 24 | Thomas Hodi

4. A more realistic View on 5G Use Cases Expectation Work Gaming Streaming Reality (sometimes) DSC DACH 24 | Thomas Hodi

5. Customer Expectations of Today 5 Rising customer expectations Need for better in-home experience OTT application experience Broadband connectivity In-home / Customer domain CPE WiFi Apps 48% customers working from home are concerned about internet stability affecting their work 1 of yearly data traffic growth in mobile networks, 20% growth in fixed 2 28% 57% of traffic from 6 largest OTTs (Google, F acebook , Netflix, Apple, Amazon, Microsoft 4 ) of traffic is video, YouTube and Netflix alone causing 27% of total downstream traffic 4) 54% Top-3 customer reasons for a new Internet-at-Home contract in A1 Austria 3 : Stable connection Reliable and Fast Wi-Fi Speed/bandwidth 1) Source: Amdocs, 2) Source: FMTR AT, 10/2022 vs 10/2021, 3) Source: 66-68% responses Brand Finance Report on Brand Value 2022 for AT based companies. 4) Source: Sandvine Internet Phenomena Report​ DSC DACH 24 | Thomas Hodi

6. 1 mobile base station 3,000 KPIs / hour 7,000 mobile base stations (Austria) 21,000,000 KPIs / hour 50,000 mobile base stations (A1 Group) 3,600,000,000 KPIs / day 1 mobile subscriber 150 KPIs / call segment 5,000,000 mobile subscribers (Austria) 750,000,000 KPIs / call segment 25,000,000 mobile subscribers (A1 Group) 375,000,000,000 KPIs / day* + 2,650,200,000,000 KPIs / week = Let’s do some Maths… * assuming 100 call segments per subscriber per day 2 Trillion, 650 Billion, 200 Million DSC DACH 24 | Thomas Hodi

7. 2,650,200,000,000 is a huge number! Traveling 2.65 trillion km would be like making 17,700 trips to the sun and back Count one number per second - it would take you around 84,000 years Spend $1 million every single day, it would take you 7,260 years to spend it all DSC DACH 24 | Thomas Hodi

8. Operating Mobile Networks is a Challenge! * assuming 100 call segments per subscriber per day 2,650,200,000,000

9. SARA Deployments We generate valuable insights from Network Data and make it easily accessible.

10. Network Analytics Product ‘SARA’ SARA Platform Service Capacity Plan Operate Care Smart Capex Config Management User Analytics Performance Management API Products Pillars Platform RF Fingerprint Energy Strategy Advisor Energy Monitoring GUI Anomaly Detection Smart Saving Device Analytics Query Engine SON Site Acceptance Proactive Care

11. Platform Components Cloud-native | Multi-cloud Works on Azure, AWS, Google, Openshift Cloud native, supporting agile CI/CD environment Open Source No PaaS components, 100% open source Features API Gateway Authentication & Authorization SQL Query Engine ( Trino ) Scheduler (Argo) Logging & Monitoring (Grafana) Rules Engine … DSC DACH 24 | Thomas Hodi

12. Data Democratization Data Exposure via API Custom reports Customer applications Hundreds of endpoints Query Engine for Big Data Analytics Super-fast data exploration Easy data visualization Setting up alerts DSC DACH 24 | Thomas Hodi

13. “World, we have a Problem” Global average land – sea temperature anomaly trend during last decades DSC DACH 24 | Thomas Hodi

14. ICT accounts for ~5% of Global Electricity Consumption The ICT footprint has doubled between 2009-2023 50% 15% but double-digit growth 15% 10% Source: The World Bank Report 2023 DSC DACH 24 | Thomas Hodi

15. A1 Telekom Austria Group | ESG Targets 2030 Target P romote the circular economy at the company: recycle around 50,000 old devices a year Target Increase energy efficiency by 80% compared to 2019 Energy efficiency index Defined as electricity consumption per terabyte of data transported 0.18 0.03 2019 2030 80% due to measures Energy Efficiency Circular Economy Collected old mobile phones (in pcs) 50,000 pcs … Energy- efficient equipment Mobile phone recycling initiative Energy- efficient network components Refurbished devices Levers Levers Target Reach net carbon neutrality by decreasing own carbon footprint, and gradually switching to energy from renewable sources Emission reduction path 2030 ( SBTi approved) CO 2 Emissions 241,000 t CO 2 120,000 t CO 2 via measures Via compensation Net carbon neutral 0 t CO 2 Electricity consumption Reduce fleet & switch to carbon neutral cars Electricity from renewables Low carbon emission heating Levers DSC DACH 24 | Thomas Hodi

16. Available vs Consumed Network Capacity Available Capacity Source: Ericsson DSC DACH 24 | Thomas Hodi

17. Smart Energy Saving | ML Features Implements multi-step hourly forecasts for site utilization. Trains XGBoost models for a one-week forecast period. Predicts hourly sector utilization for the upcoming week. Reports error metrics (normalized RMSE, MAE) over two weeks. Calculates confidence intervals using the percentiles method. Planned: Incorporate weather data to improve prediction accuracy. 1010011010010000101010011110111011011011010101000011100101011001010100111010100010101000101101011011011010001010111000101010001010001011101011000100110100110100100001010100111101110110110110101010000111001010110010101001110101000101010001 BIG DATA DSC DACH 24 | Thomas Hodi

18. Forecasting Algorithms Example 1 “Easy-to-model” Site Example 2 “Problematic-to-model” Site Forecasting Algorithm | Accuracy Examples

19. ML: Slice of an Expert System DSC DACH 24 | Thomas Hodi

20. Anomaly Detection Analysis of network performance KPIs to identify anomalies = deviations from normal patterns. Algorithm: LSTM Autoencoder model to handle time-series data, detecting anomalies based on historical patterns. Process: Reads hourly KPI data, applies pre-trained models, calculates anomaly scores, and outputs results to storage systems. Value: Faster and more targeted root cause analysis. DSC DACH 24 | Thomas Hodi

21. ML: Slice of an Expert System (again!) DSC DACH 24 | Thomas Hodi

22. SARA Deployments Austria Slovenia Croatia Serbia Bulgaria North-Macedonia Brazil Colombia (under discussion) DSC DACH 24 | Thomas Hodi

23. Our Achievements Source: MWC Americas 2019 DSC DACH 24 | Thomas Hodi

24. Challenges & Learnings how to Operationalize AI/ML Data Skills Scaling 56% 55% 20% of CSPs are facing data quality issues lack the right data science skills of AI PoCs progressed to live deployments Source: 2019 Nokia study of 50 CSPs  “A1 Group” Data Lake  Partnering  Building a Platform DSC DACH 24 | Thomas Hodi

25. human@center DSC DACH 24 | Thomas Hodi

26. Interested? @ [email protected] linkedin.com/in/ thomas-hodi /