Brügge Kai

Abstract Reference: 30830
Identifier: P1.5
Presentation: Poster presentation
Key Theme: 1 Reduction and Analysis Algorithms for Large Databases and Vice-versa

Real Time Streaming Analysis of IACT Data

Authors:
Brügge Kai

Imaging Atmospheric Cherenkov Telescopes (IACT) like FACT produce a continuous data flow during measurements. The First G-APD Cherenkov telescope (FACT) is dedicated to monitoring bright TeV Blazars in the northern sky. FACT's continuous monitoring produces a dense sample of observation points over long periods of time. To help understand the mechanisms of cosmic ray acceleration, the sources need to be observed over a wide range of wavelengths simultaneously. To coordinate successful multi-wavelength campaigns, other experiments need to be alerted quickly in case of flaring sources. This puts strong constraints on a real time data analysis process. It needs to analyze the data faster than the telescope writes it. Otherwise, delays will accumulate over time. The usage of silicon photon detectors in FACT allows for a larger duty cycle compared to traditional Photo Multiplier Tubes. This results in large amounts of collected data. Given good weather conditions, up to 1 TB of raw data are recorded per night. To calculate excess rates from raw data, the entire data analysis process including data calibration, filtering, image cleaning, parametrization and signal/background separation needs to be performed. The online analysis software for FACT is build on top of the streams-framework, a modular data streaming environment working in conjunction with popular Big Data solutions for distributed computing like Apache Spark. Pre-trained multivariate models are applied to the data written by the telescope data acquisition system. This allows for effective background suppression in real time. The system features a web interface showing live analysis results and telescope status during measurements. We show excess rates from observations of two bright blazars Markarian 501 and Markarian 421 and cross-check the results with an existing Quick-Look-Analysis for flare alerts. The technologies demonstrated here also serve as a testbed for the upcoming generations of IACTs as part of the CTA project.