We decompress that, process it, and ship out 15 to 20 petabytes of imagery per year to our customers. Most recently, we're putting it into our geospatial cloud to make it available on a much more real-time basis.
What's involved in processing that data? We stretch the images across an elevation model of the Earth to make sure that they're accurate. We cancel out the angle of the camera that the image was taken from; we cancel out the bumpiness of the Earth; we correct for the wobble associated with the gravitational pull of the moon and what that pull means to the tides. All of those things need to be corrected. It's a massive computational problem to create accurate imagery.
What role does analytics play in your business and IT operations? We use analytics to keep our images current, to determine where we need to refresh next. Cities with a high rate of change are high on this list because our customers use our imagery to reset car navigation systems.
What unique challenges does IT face in meeting the needs of your industry and business? It's the big-data problem. We process massive amounts of imagery, and we need to meet very exact timelines. So the challenge is how to create a flexible computing model that lets us create the right product for the customer. We created our own high-performance compute cluster, and we are actively harnessing the power of graphics processing units to do this. By moving processing from CPUs to GPUs, we have seen anywhere from a 10- to 20-times improvement in speed.
Today we have somewhere around 20 petabytes of data between spinning disk and tape. In the future, customers will no longer be able to [receive] all of the data they need to answer a problem. The data is becoming an immovable object. So we're going to create in our cloud an opportunity for customers to bring us their computations, which we can then run on our high-performance computing environment. It will be easier to bring the computations to the data than to bring the data to the computations.
Can you discuss some of the mobile projects you're exploring? We have a customer that provides location-based services through an online navigation portal, and they want to make our images accessible to their mobile customers. If you're looking at imagery through a phone or online, it comes back in little squares, or tiles. To create this ability to tile out the entire globe and serve it to mobile customers requires somewhere in the neighborhood of 42 billion tiles. We put that into our high-performance computing cluster, and we are in the process of finishing up delivery of those 42 billion tiles.