Real Time Analytics for Measured Spectrum Data NSF

Real Time Analytics for Measured Spectrum Data NSF Spectrum Measurements Workshop IIT, Chicago, IL April 6 th, 2016 Joydeep Acharya Global Center for Center Innovation Hitachi America Ltd. , Santa Clara, CA 1

Introduction § What is spectrum data? v Power levels at scanned frequency channels o Sophisticated wideband sensors, narrow resolution BW, frequent reporting o Generate huge volume of data (~TB/hour) v Information about the actual signals o Type of modulation , protocol analysis v Non traditional sources – device measurements, crowdsourcing, wireless carrier data v Rich variety of data to be analyzed into actionable insight § What is spectrum analytics? v Spectrum occupancy reports – primary vs secondary usage v Spectrum interference & misuse reporting and identification v One time values, spatio-temporal trends 2

Data Analytics for Spectrum § Plethora of Database management technologies ‒ Data at Rest (traditional applications) ‒ Ingest -> Store -> Index -> Analyze ‒ Structured to unstructured data, batch processing, graph representation etc. ‒ Data at Motion (real-time, streaming applications) ‒ Analyze data as it is ingested (in memory processing) ‒ Low bandwidth requirements and latency ‒ Structured to unstructured data and micro-batch processing or pure streaming analytics § Real time processing may be suitable for advanced spectrum analytics 3 ‒ Ability to join with historical data (data at rest) for trend analysis ‒ Integration with user defined functions (R, Matlab) for machine learning (anomaly detection) ‒ Edge processing and adaptive ingestion to reduce centralized load