The Theoretical Astrophysical Observatory (TAO) houses queryable data from multiple cosmological dark matter numerical simulations and galaxy formation models, in a database optimised for rapid access. TAO boasts a clean and intuitive web interface. It avoids the need to have the user program SQL queries by providing a custom point-and-click interface for selecting galaxies and their properties, which auto-generates the query code in the background. Query results can be funnelled through additional “modules” and sent to the Swinburne OzSTAR supercomputer for further processing and manipulation. These include the ability to:

  • Construct observer light-cones from the simulated data cube;
  • Generate complete spectral energy distributions for model galaxies to provide multi-wavelength coverage;
  • Produce mock images of the sky in each simulated universe;
  • Select from preset “virtual telescopes” that make reconstructing existing surveys easy.

It uses a cloud-based architecture that offers Infrastructure as a service (IaaS), Data as a service (DaaS), and Software as a service (SaaS). It is configured in such a way that additional modules can be built and easily inserted into the data chain as needs require in the future. Refer to Bernyk et al., The Astrophysical Journal Supplement Series 223.1 (2016) for additional details.

The development of TAO V1.0 started mid-2012 by NeCTAR funding under the Virtual Labs project. The first version was released on the 24th March 2014. The first version of the node was co-developed by Swinburne University of Technology and Intersect Australia Limited. The following list summarises the different major releases and the main features delivered within each release:

  • TAO Version 1.0 (March 2014)
  •  TAO Version 2.0 (September 2015)
    • Support for premade catalogues added, with 6 new pre-made galaxy and quasar catalogue sets initially available.
    • Enhancements to the SED module including In-place visualisation of the SED bandpass filters and the ability to download the SED bandpass filter files.
    • Significant enhancements to the Image module.
    • Support for Digital Object Identifier (DOI) data citing.
    • Integration with a JIRA HelpDesk to provide a better support experience.
    • Major UI and documentation improvement.
  • TAO Version 3.0 (November 2016)
    • Support for Hydro simulation data.
    • Support for multiple versions of the same dataset.
    • Major UI improvement (moving TAO to use React.JS).
    • A significant boost to the number of datasets available.
  • TAO Version 4.0 (expected 30th June 2017)
    • Introducing the concept of “teams” within TAO as a mechanism for grouping users.
    • Introducing private TAO datasets
    • Introducing the ability to share mock catalogue associated comments.
  • TAO Version 5.0 (December 2019)
    • Major re-architected backend
    • Support for Multiple TAO nodes (Currently Swinburne OzSTAR and NCI Gadi)

TAO architecture is partitioned into three tiers:

  • TAO User Interface – User facing component of the system.
  • TAO Workflow – Handle the integration between the UI and the back-end science modules and databases.
  • TAO Science modules – the back-end components of TAO. It is responsible for all the post-processing functionalities delivered by TAO.

The main TAO database, science modules, and results storage space are located on the OzSTAR supercomputer at Swinburne University. The user interface and Table Access Protocol (TAP) server are hosted on a separate web server and accept job requests. Jobs are queued on the supercomputer via the jobs scheduler. Upon the job completion results are available for download over the internet.

The technology currently used:

  • Python Django (for Server-Side UI)
  • React.JS (for Client-Side UI)
  • C++11 for Science modules implementation
  • OpenMPI for Science module distributed implementation
  • Python 3.6 for the Workflow module and data importing
  • PostgreSQL Databases

Hardware arrangement:

  • Two active-passive load balanced nodes for TAO UI hosting
  • 3 Database nodes (2 Intel Xeon E5-2660, 64 GB RAM)
  • 100 TB of Lustre storage
  • Unrestricted Access to the SwinSTAR cluster (main processing)
  • QLogic QDR InfiniBand with 3GB/s per node non-blocking communication
  • Two virtual nodes for the support and documentation hosting.

TAO is the only node at the moment that provides professional grade support (via asvo.atlassian.net with less than 48 hours SLA) and externally hosted documentation and tutorials.
The primary stakeholders:

  • Centre for Astrophysics and Supercomputing – Swinburne University of Technology
  • Swinburne HPC