There's a "lump in the learning curve" for some organisations to overcome in the adoption and use of NoSQL databases. That's what James Tomkins, Met Office Portfolio Technical Lead, told Computing while discussing the organisation's use of big data.
Working with partners including US space agency NASA and the UK Space Agency, the Met Office is using a MongoDB-powered NoSQL database to predict space weather events including solar flares and coronal mass ejections.
Further reading Analysis: It takes skills to explore big data Analysis: The inexorable rise of NoSQL NetApp helps CERN to crack riddle of the universe
The aim is to provide better information about astronomical phenomenon to governments and businesses so that they can take appropriate action to minimise the impacts.
Last year, for example, the biggest solar flare in five years hit Earth, which meant that aircraft had to be diverted away from polar ice caps in order to minimise the risk of communications outages during flights.
However, while big data isn't just for the likes of the Met Office and NASA, there are challenges to overcome in its adoption.
"In terms of non-relational data structure, I think there's a definitely a lump in the learning curve for people to take their first steps away from the more traditional, more well-known relational data-model handling," Tomkins told Computing.
He added that despite some barriers, NoSQL databases provide definite benefits.
"But I think the more and more you look at data, the more occurrences you see of data that isn't traditionally relational and you do want to store and interrogate in ways that don't lend themselves to more traditional structure," he continued.
Tomkins added that sticking to traditional methods, however, has drawbacks and that for the Met Office, the previous requirement for sticking with one representation of its data has probably held it back.
"I think there will always be a place for relational data, but I think that as we look more and more around the organisation now, we're identifying requirements where, perhaps, we're not modelling and storing and using our data as best we can, because we've had to always stick with the one representation of our data," he said.
