The Virtual Universe: The future of astronomical computing What will astronomy be like at the end of the 21st century? In my vision, instead of publishing your results on astro-ph (and, archaically, in the Ap.J), you will submit data files and code subroutines to the Virtual Universe. The `data file' part of this is already close: an archive which will accept heterogeneous data. The trick will be to make all such data sufficiently self describing that it can fit into the VU's object model. Among other things, it implies that the VU "observational data" objects will have attributes including 'degree of rawness vs calibration', 'degree of refereeing' (unrefereed, minimally reviewed, or the equivalent of ApJ refereeing), and sophisticated descriptions of data quality and limitations. The `code' part is further away, but we can start to see how to do it. Today we write simulation and analysis code, often mixed together, for specific projects and then usually throw the code away. One person's N-body simulation is not compatible with another's, and it's hard to compare them, and almost impossible to put the output in the input of someone else's observational simulator. It's a lot more work to define standards and conventions so that a higher level program can stick these pieces together, but I believe it must and will happen in the long term, that this will become a requirement of publication. In that utopian day we'll be able finally to say: "Computer! Run a stellar population synthesis using Smith's IMF and Lee's gas dynamics treatment in Popov's elliptical galaxy potential, and observe the result for 100 ksec with Virtual NGST for a range of galaxy orientations, masses and redshifts", and have the computer actually obey you. This implies an object model for the universe with hooks for many different levels of treatment (e.g. extinction laws would range from simple power laws to full radiative transfer with physical dust grain and dust spatial distribution models). An important part of the Virtual Universe will be the Virtual Observatory, in which (real) astronomers carry out (virtual or real) observations and (real) data analysis. (The introduction of Virtual Astronomers is deprecated in the interests of job security). The operation of virtual observing implies applying a Virtual Telescope to some part of the Virtual Universe. In many cases this involves creating or accessing the Virtual Sky, a two-dimensional spherical projection of the four (or possibly 11) dimensional Virtual Universe (but we should also support in situ measurements). The output of the Virtual Telescope is a Dataset, as currently supported by the Chandra DM. Archived observations of the real sky become part of the Virtual Sky, as do some simulation data. We should also support the special case of remote observing, when the True Sky is accessed with a real telescope in real time. (Note that the True Sky is read-only, at least at current funding levels). The astronomer accessing the Virtual Observatory will use command line or browser interfaces to construct an Astronomy Query. The query object will define an Astronomical Workset, comprising zero or more Datasets (either retrieved directly or constructed by applying on-the-fly analysis to other Datasets) and zero or more Worksheets, which are display pages which might be implemented as HTML pages. In simpler language, the user may want simply to get a display page giving some summary information, or may want to get their hands on some actual data, or both. The structure of the VU is then ------------- Virtual Universe -------------------------------------------- VU Physics VU Astronomical Objects V-Telescopes (code) (simulations, models) (simulators, calibration data) | | | ------------------------ ----------------- | | V On-the-fly Datasets | Real-time Datasets-->V-Sky<---- Archived Datasets | | | | | | ----------------- VO Portal ------------------------------------------ ----------------- User Interface ------------------------------------------ | | ^ v | Workset | Astronomy Query The VU is of course distributed, and the job of the VO Portal is to coordinate requests and translate them to the language of the different data providers as discussed in the NVO concept studies.