Abstract

A High Performance Spatial Query Engine for Large Data Sets Implemented for the GLAST LAT Data

Thomas Stephens (Wyle/NASA GSFC)

GLAST Science Support Center

The Large Area Telescope (LAT) is the primary instrument on-board the Gamma-ray Large Area Space Telescope (GLAST) satellite. The LAT is not a typical imaging instrument but rather records the energy and arrival direction of each individual gamma-ray photon. It is designed to operate primarily in sky survey mode, instead of pointing at fixed targets for long periods of time. The standard survey mode gives complete coverage of the celestial sphere every 2 orbits (~3 hrs). Additionally, the LAT has a very large (>2 sr) field of view, and an energy dependent point spread function (PSF). These factors combine to generate a large data volume and present a unique challenge in providing data to the user community for the study of astronomical sources. We present the design of the public data server at the GLAST Science Support Center (GSSC) for the LAT data as well as performance benchmarks of the currently operating server. We discuss the hardware and software choices made to allow the system to achive its high performance. Based on the user's query, the photons matching the data cuts are extracted and presented to the user as a FITS file ready to be downloaded and used in the GLAST science analysis software. Running on a single CPU, the photon data server can extract and prepare the ~2-3 million photons matching a typical user's query from the ~140 million photons in a year's worth of observational data on the order of a minute. The complete system is implemented as a small cluster of Linux PCs to improve performance even more by distributing the load of a single query or processing multiple queries simultaneously.

Mode of presentation: oral