Barclays aims to reduce its reliance on ageing mainframe systems by moving its customer data onto a more agile and resilient NoSQL-based data platform.
Like many large banks, Barclays has faced a significant challenge in meeting customer demand for new digital banking services while relying on legacy infrastructure. It is a topic that has drawn the ire of regulators in the past due to outages which have affected customers. Read next: FCA to assess banks' legacy IT systems as outages continue
“As with a lot of large institutions, most of our critical applications are on mainframes,” said Bala Chandrasekaran, Barclays platform director, speaking at MongoDB Europe in London today. “That creates a single point of failure.”
Last year, the bank experienced two service outages as a result of mainframe failures, which prevented customers from making payments.
To help prevent further occurrences, the bank investigated how it could access customer data when mainframes systems became unavailable. This resulted in the creation of an operational data store (ODS) based on MongoDB’s NoSQL database software, that sits between services such as its online or mobile banking and its mainframe systems.
Barclays staff had discovered that that the vast majority of its customer service demands centred around accessing ‘read only’ data, such as account balances.
"That led us to an interesting question: If most people are only reading data, is it possible to keep a snapshot of the data that is sat in the mainframes as a 'read-only' copy and make it available to channels when the mainframe goes down?
"So, rather than stop everything and prevent customers from being able to do anything with us, why not allow customers to be able to get their balance and transaction - which is what they would simply be doing most of the time?
"That is why the journey with ODS started."
The ODS platform went live earlier this year, and is already used in a variety of ways within the bank.
“One of the big ones is transaction history,” said Chandrasekaran.
In the past, online transaction histories were limited to 300 entries, due to the difficulty in retrieving the huge amount of data from the mainframe systems. As well as being problematic for personal account customers, this was an issue for small businesses which could have hundreds of transactions in a relatively short period of time.
“You could see only a subset [of transactions] because we couldn't store and serve up that much,” he said.
“It was just a capacity issue, so the first use case we had with the MongoDB ODS was to bring all of the transaction history data and make it available to the [online and mobile] channels."
Using ODS as a caching platform, it was possible to retrieve large volumes of transactions almost immediately.
“The ODS is literally in near real-time sync with the mainframe,” he said.
“So if you login and you say 'I want to see my transactions that has the word ATM in it for the last two and a half years', it is able to serve it up in 600 milliseconds. This is an actual use case which is live today.”
Reducing reliance on mainframes
Chandrasekaran said that the bank is quickly expanding the use of the ODS platform. This means using it as the “first port of call” for customer data, rather than just as backup for mainframe systems as had initially been intended.
One example he gave was serving up data for online and mobile banking ‘landing pages’. This is one of the most intensive data procedures that the bank will have to perform - querying personal accounts, savings and mortgage databases at once, for instance - and occurs every time a customer logs in to access digital banking services.
“We are hitting the mainframe five million times a day for one of the heaviest queries,” he said.
Instead, Barclays plans to hold the customer data in the ODS platform, allowing it to be retrieved in close to real-time when a customer reaches the landing page.
“[With MongoDB] It is a single lookup and that makes it faster, simpler and extremely cheaper.”
This has meant a departure from the initial aim of essentially providing back up to mainframes, and instead benefiting from the advantages of the NoSQL technology.
He said: “What we started off as a resilience use case we are now also making a mainframe offload use case. This is currently in progress and we are hoping to go live by Q2 next year.”
Personalisation and ‘open data’
Going forward Barclays plans to expand the use of the ODS platform as it aims to provide greater personalisation in online customer services.
“One of the trends now is it is no longer just about pushing generic offers to people,” Chandrasekaran said. “When you log into our [online or mobile] channel there is an offer that is specifically customised to you based on what you have done and we are looking to use the ODS as a vehicle to deliver it.”
In addition, with the PSD2 (payments systems directive) on its way, the NoSQL platform will help open up the bank’s data for new business opportunities.
One area of potential is to enable customers to view their account balance via third party websites such as Facebook.
“It is conceivable that you no longer log in to Barclays or Santander or any other bank, you just go to Facebook and say 'show me my balance’,” he said. “Facebook goes to all these banks, brings up data and it shows you. You say you need to buy this cool gadget and Facebook will say there is a very good offer in Barclays, so use your Barclays account to make the payment.”
He added: “It means we need to be able to scale up elastically. So the PDS2 is one of the key things we are pursuing under the ODS.”