Met Office ditches Google App Engine for Microsoft Azure to 'future proof' Weather Observations Website

Azure's PaaS model a ray of sunshine for crowdsourced weather database

The Met Office has decided to move from Google's PaaS (platform as a service) cloud hosting and development service Google App Engine to Microsoft's Azure platform, in order to "future proof" its Weather Observations Website (WOW) for the coming Internet of Things (IoT) age.

The WOW initiative is an example of the big data-led projects the Met Office is now focusing on, as recently outlined to us during our interview with acting CISO Tim Moorey.

Speaking to Computing about the move to Azure, CIO technical adviser Mark Burgoyne explained how the project behind WOW - the Met Office's crowdsourced site for public weather observation submissions - is keeping its "original aim" to gather data from the "weather enthusiast community", but is now looking to extend its reach in order to integrate data from mobile phones and even cars.

"The Google App Engine platform at the time was pretty much the only solution that allowed you to have that PaaS programmable functionality - they were the only major vendor out there," Burgoyne told Computing.

"Other people were offering infrastructure as a service (IaaS), but we really wanted some sort of PaaS in order to lower the overhead without any requirement for patching - you're just running your software in something that automatically scales up."

In 2015, as a part of a wider project to "refresh the look and feel of the site", the WOW team decided to look for enhanced cloud functionality and a cloud platform provider that could extend the scope of data "in an affordable and scalable manner".

"There's a community out there generating the same sort of data we've always collected from our own formal network - but it's now a lot denser and there's a lot more of it," said Burgoyne.

"We also want to be able to cope with the ever-growing variety of new sources of data. There are lots more devices out there that are recording information that might be useful to us - things like temperature readings from cars or pressure sensors in phones, so we wanted a platform to take account of that," he added.

Greater flexibility was also a factor. "We wanted to be able to easily configure new types of data, and have that built into the platform," said Burgoyne.

"There's other options now as well, going beyond just PaaS - you've got SaaS options in there. So the Azure platform solution came along and fitted quite nicely with that combination of PaaS - where we or our partners are writing their own logic modules, and reusing software functionality, all scalable - really reducing our management overhead.

"Looking at the various options that came along, Azure was just the match for what we were trying to do."

Maria Dorothy, project manager for observations, added that since 2011, there have been over 850 million observations sent to the Met Office.

"That's an average of 20-30 million observations per month, and these are now increasing month by month, and this is just from specialist weather station data alone, before the addition of IoT-enabled everyday devices," said Dorothy.

The data also includes the venerable "0900" service that lets hobbyists phone in their daily observations.

"The main aim here is to try and future-proof ourselves," said Burgoyne, "but it needs to be relatively easy for us to add new types of information quite quickly."

Azure Machine Learning and Power BI are also on the table for the Met Office and WOW.

"That's the next stage - once we have a large volume of information, those sorts of tools offer us some really interesting possibilities for things like identifying station quality, and the quality control side generally," said Burgoyne.

"The large amount of variables involved, and how easily they can be integrated, is one of the things we like about the Azure platform.

"The Google App Engine we used before was something you wrote software for yourself on, and it could scale your code. [But] I think Azure, and its PaaS model, offers lots of off-the-shelf functionality we could bolt on as and when we need it, which is where the cloud benefit comes from."