Flexible Service Composition for Astronomy

Breakthroughs in telescope, detector, and computer technology have allowed astronomical instruments to produce terabytes of images and catalogs. The data sets produced cover the sky in multiple band widths,from gamma and X Ray, optical, infrared through to radio. With such vast quantities of data being archived, it is becoming easier to "dial up" a piece of the sky, rather than waiting for expensive, scarce telescope time. The software which allows the integration of astronomical resources has been slow to catch up with the ever increasing astronomy data volumes. Virtual Observatories are the technology frameworks which aim to fill this gap, allowing transparent access to astronomical archives, databases, analysis tools and computational services.

One way to support experiments in virtual observatories is via workflows describing experiment plans, where the primitive steps in the workflows correspond to data provision, curation, transformation and analysis services. The Astrogrid project produced its own system for specifying and managing these workflows but this language operated at a comparatively low level of specification (compared to OpenKnowledge) and required engineers to pay attention to details of service invocation and orchestration for every new workflow. As an experiment in applying OpenKnowledge methods to this domain, Barker took a variant of LCC (the MAP language developed by Walton) and used it to specify workflows used in Astrogrid. For a more detailed description of this effort, see Barker & Walton's report

The Large Synoptic Survey Telescope (LSST) project pushes beyond conventional workflow analysis and is intended to open up "time domain" astronomy; the telescope will be able to tile the entire night sky over a three night period, generating 36 gigabytes of data every 30 seconds. The data reduction and analysis in LSST will be done in a way unlike that of most observing programmes. The data from each image will be analysed and new sources detected before the exposure for the next tile is ready. This means that if anything unusual is detected, normal observation can be interrupted, in order to follow up any new or rapidly varying events. The hope is that, since LSST will provide a first attempt at time domain astronomy, it will discover new classes of object, previously undetected. Once the initial processing has finished, there will be some data which is left over. This data includes objects and orbits which can't be classified by the processing software. Many of these objects will simply be junk, but this may only be revealed on the basis of comparison with other detections made from the same night. The systems which attempt to classify this data need to be reactive, collaborative systems. Although the LSST is still under construction, Barker and Mann have used the MAP variant of LCC to describe a system of coordination that could be used between agents interpretign LSST data. For a more detailed description of the LSST scenario, see Barker & Mann's report For more details on MAP applications contact Adam Barker. For backgrond on astronomy contact Bob Mann.