The Locus Algorithm: The design, implementation and performance characterisation of a software and grid computing system to optimise the quality of fields of view for differential photometry

Abstract

We present the development and software implementation of the Locus Algorithm, a novel algorithm designed to maximise the performance of differential photometry analyses by optimising the number and quality of reference stars in the Field of View with the target. Firstly, we state the design requirements, constraints and ambitions of the system. Initial testing showed the necessity for a High Performance Computing system such that the software system could be scaled to analysis of the full SDSS Catalogue in a practical timescale. The National Grid Infrastructure at Grid Ireland was selected to fulfil this need. Then, a detailed software design is presented for the system in operation. Next, the data design including file structures used and the data environment required for the system are defined. We present an analysis of the performance characteristics of this system, including an assessment of the scalability of the solution. We conclude that the balance in our system between I/O and processing time varies between I/O dominant and balanced depending on the inputs. Therefore, we note that for our future projects, such as applying this analysis to larger datasets such as LSST or Gaia, we may not be able to rely solely on factors such as algorithmic complexity when predicting system requirements. Instead, experimental testing of runtime and careful analysis of optimisations such as data reuse are critical.