Breaking the boundaries, saving lives: How cloud can prevent future crisis
In our last case study we talked about how the use of cloud in cancer therapy can push the accuracy of treatment for patients to near real-time. The use of cloud on a highly personal level shows the amount of detail possible by accessing large-scale cloud resource for short periods of time. But what if we were to go in the opposite direction? What if the potential crisis impacted hundreds, if not thousands, of people? What could cloud do when the fate of whole cities is on the line?
The experts at Sterling Geo were contacted by the World Bank with just such a challenge: to harness the cloud for a collaborative mass mapping programme. They were specifically tasked with a large-scale unmanned aerial vehicle (UAV) image processing project for East Africa. In layman’s terms: to process thousands of images taken from drones to compile into a map targeted at flood prevention. Their primary target? Dar es Salaam, a city that is growing at a rapid pace where there is a dearth of data to be processed into actionable information for city planners and developers.
The only problem? A UAV imaging project of this size had never before been attempted.
To get to the goal set by the World Bank, Sterling Geo commenced with a phased project process. The first phase was to understand how the cloud could be harnessed to process hundreds of thousands of UAV images rather than the traditional standalone workstation approach. Historically, the use of data processing has always been restricted to the size and capacity of fixed servers and, in most cases, to the size of the workstation available to the UAV imaging expert. What happens when those restrictions are no longer an issue?
The team at Sterling Geo contacted Alces Flight to take the technology developed by their experts and explore how it could be adapted for repeatability and scale by using high performance computing (HPC) in the cloud. The imaging technology, which the Sterling Geo team wrapped into a container for preservation of IP and ease of delivery, was connected to Alces Flight Compute to create a repeatable, auto-scaling mechanism for image processing. In the field, a UAV pilot uses this template to start a HPC cluster on demand, allowing them to immediately upload their images and commence processing. Once the results are ready, the HPC cluster shuts down. Not only does the template launch and control the HPC cluster, so that resources are used responsibly, it also ensures that imaging results stay similar across all the pilots in the field.
Initial results from the project are yielding rapid progress in how the transition from workstation to cloud can be managed and demonstrating how scaling to cloud realistically delivers vast improvements for image processing.