Wikipedia Explorer case study
The client: Microsoft has invested many billions of dollars into its new Windows Azure platform and was looking for a killer application to show its power.
The challenge: Wikipedia Explorer is all about visualising relationships between documents within Wikipedia, in an attempt to improve the traditional, flat user experience. In the original version, all the data was downloaded from Wikipedia on the fly. This was very simple, but also very slow and meant different visualisations would be produced each time, based on which links were downloaded first. Microsoft wanted a much faster browsing experience – with much more continuity – created in Windows Presentation Foundation.
The solution: Dot Net worked alongside Monochrome and the Microsoft developer and platform evangelism team in Redmond, Washington, to deliver a robust proof of concept for this new browsing experience.
Wikipedia makes available a database dump, which contains a complete snapshot of the whole site. However, the problem is the dump is enormous. Converting from wikicode (the proprietary format it is stored in) to XML/XAML is very processor intensive – on a single high-powered server it would take somewhere in the region of 4-6 months to complete. Obviously, this was a non-starter.
By building the application on top of Windows Azure we could scale this process out to a large number of servers. It currently runs on 50 server instances. The same process that would have taken up to six months took a little over four days – almost exactly 1/50th the time – demonstrating the power of the Azure Services platform. It is easy to scale out processor intensive tasks and have them completed much faster by provisioning more hardware.
Wikipedia Explorer is a regular in the keynotes of conferences across the UK and America.