How Barclays Avoids Downtime Chaos

Current consumer technology trends such as eCommerce, ever more powerful smartphones and contactless payment technologies are providing traditional retail banks with unique challenges in order to remain competitive. Those same banks are also being required to control ever escalating costs.

Barclays has been a pioneer in Digital Self Service capability roll-out. In addition to its feature-rich online banking platform, in recent years Barclays launched Barclays Mobile Banking (BMB) and Barclays Pingit, which supports Barclays and open market customers.

Adoption and usage of its mobile banking apps have exploded in the past three-plus years since their launch, with more than 6.5 million active customers using them. These customers, on average, log on to their accounts via mobile app 26 times per month, and Barclays experiences five million logins per day with 800 requests per second at peak times.

No Downtime

“The data on usage is very clear. People tend to login on their way to work, at lunchtime or on their way home from work,” says Bala Chandrasekaran, Barclays director of data optimisation simplification says. “People have stopped coming to their local branch or calling us up as much, which in some ways is great because people were finding it very easy to use but it created a very interesting problem. Sometimes too much of a good thing can be a bad thing.”

Like most organisations offering banking services, most of Barclays’ critical apps run on the mainframe, which creates a single point of failure.

“If the mainframe goes down, which it does more often than you think it should, or more importantly the connectivity to the mainframe goes down, it brings the whole service down,” adds Chandrasekaran.

“When this happens, it causes significant impact on channels and customer experience. There’s no way for us to service our customers and that’s when you end up making front page headlines in the newspapers.”

On the back of a couple of such incidents in Q4 2015, a Barclays task force was set up to look at options around resilience.

“The downtime we experienced was not a lot. On one occasion we were down for about an hour and on another occasion we were down for almost an hour, but it was long enough to cause chaos and this was in the tun up to Christmas,” continues Chandrasekaran.

An in-depth study came back with some interesting, though not surprising, findings in usage: As much as 92 per cent of the traffic in channels (self-service and assisted) was triggered by less than 25 distinct transaction types, and 85 per cent of those were ‘read-only’ – customers seeking information. ‘Get me balance’ and ‘show my transactions’ were the top transaction requests two across all channels.

Chandrasekaran notes: “Even when people phone up our call centres, more often than not, they want to request information rather than actually complete transactions.

“That led us to an interesting question. If most of our customers only want to access data, is it possible for us to keep a snapshot of that data sitting on the mainframe as a read-only copy and make it available to channels when the mainframe goes down?”

What happens when the mainframe goes down? More on Page 2…