UBS has moved its regulatory data onto a NoSQL database platform, allowing the bank to explore archived information for business intelligence purposes.
Like all large banks, the Swiss giant is required to gather an increasingly large amount of data in response to the demands of new regulation in the sector.
However, USB intends to make use of this customer transactional information - that could otherwise be left largely untouched for years - by accessing it through a non-relational database. The bank had previously relied on 60 relational database applications to hold its regulatory data.
"We have the sets of data there, so we are looking at what different purposes that we can put this to," said UBS head of E & A comms IT, Mark Grant. "It goes beyond some of the original thoughts that we had."
He added: "We built this for regulation and compliance purposes, but there are other potentials to mine the data for business intelligence reasons, such as finding out commonalities between the way customers interact with us.
"It is about giving yourself the opportunity to use that data in different ways by putting it into a NoSQL database - that is really powerful."
Speaking to delegates at the 'Forrester Forum for Technology Management Leaders' in London last week, Grant said that moving to NoSQL technology supplied by MarkLogic required a shift in the way company handles its data, with a change in the way it approaches storage and querying of this information.
"Rather than the old way of thinking around archiving and retention, which was to fit things into taxonomies and to try to almost build the question we thought we needed to into the structure of the data as we were storing it, with this technology you can leave that until later," he said.
"You can revisit that data as many times as you want, so you can focus on getting that information out and getting it somewhere, and then you can start to think about the questions that you can answer with the data."
Storage cost reduction
The data storage strategy has also opened up the possibility to reduce costs involved with holding transactional information over long periods. This can be achieved through a storage tiering strategy, said Grant.
"In some cases we need to respond to queries weekly or monthly - it is a very regular use of data. Other times some data is kept for 10 years and we never need to refer to it. So we have started to look at tiering for some of the storage.
"Our current implementation has around 300TB in it. I don't need all of that online and available at the push of a button, but I do need some of it available. So what we are looking at is splitting where the data is across multiple tiers, so I can keep what I don't use as much on the lower cost, and the stuff we need more frequently on the higher level - and we can move that dynamically as well," he said.
Sign up for CIO Asia eNewsletters.