Computing the Science of the Skies

Good technology conferences have interesting content, sessions that keep you learning new things. But the best thing about them isn’t the content, it’s the people you meet and the conversations you have. NVIDIA’s GPU Technology Conference in San Jose is one of those, and today’s interesting conversation was with Jonathan Fay, the principal developer of Microsoft’s World Wide Telescope.

Mixing a desktop rendering engine with multiple data sources and cloud services, Microsoft’s World Wide Telescope is one of company’s triumphs, and a fitting memorial to Jim Gray (the database pioneer who built the original Sky Server). It’s a window into a terapixel view of the known universe, across a wide range of servers and offering views of the sky at multiple wavelengths. It’s a fascinating tool, and well worth spending some time exploring the huge database of images and mapping information, travelling backwards and forwards in time.

Talking to Fay, it’s interesting to see just how accurate the simulations that drive the WWT are. The solar system model contains over 500,000 separate objects, modelling as many of the asteroids and comets as possible. It’s a model that needs GPU processing to work, as CPU-based techniques wouldn’t able to handle that many objects in real time. It was those asteroids that started the conversation, as it surprised me just how spread across the sky the leading and trailing Jupiter Trojans were…

Fay showed us how the model handles the orientation of the ISS, noting that if you looked up at the sky with a telescope, the station would be oriented the same as in the WWT model. It’s a detailed 3D model, too, with support for stereo rendering, giving a clear view of the station (and in Fay’s example, the slowly separating shuttle).

But that’s not the only use of GPU technology. There’s also support for rendering interactions between different bodies. As Fay rotated the globe, we spotted a small black spot in the Pacific Ocean. It wasn’t a bug in the simulation – this was a rendering of the shadow from this year’s major eclipse. Fay then flipped the view, dropping us down onto Easter Island where we could watch the moon slowly cover the sun, showing the corona in the moment of totality. It wasn’t quite as good as being there, but it was a start.

WWT isn’t just a tool for viewing the sky. If you’ve got geographic information you can take it and render it on WWT’s NASA sourced earth imagery. Fay was demonstrating how an Excel spreadsheet of earthquake energy releases in the Caribbean showed many, many small quakes along the main fault lines, with none near Haiti – until the massive quake that caused so much devastation. It’s a technique that lets you quickly visualise science on a large scale.

That’s been one of the themes at this GTC, the rise of computational science and the use of GPGPU computation to bring it out of the supercomputer and on to the workstation. We tend to forget the power of analytical science – something that goes back to the early days of astronomy, where Kepler’s analysis of Tycho Brahe’s observations led directly to an understanding of the laws of planetary motion. Computational science means using brute force processing to analyse large amounts of data, and it’s not really surprising that the next generation of NVIDIA’s GPGPU silicon will be called Kepler…

Simon Bisson