How to get mainframe modernisation right

Around a third of modernisation projects that lift and shift mainframe workloads to a distributed architecture often fail, according to a regional executive at Rocket Software.

In an interview with Computer Weekly, Praveen Kumar, Rocket’s vice-president for Asia-Pacific, singled out a few factors that have been driving organisations off the mainframe: the lack of skills to manage mainframe workloads, the rising costs of doing so and the slower speed to market.

“All three factors are interlinked, because if you have fewer skills, your costs go up, giving you fewer resources. Every time there’s a change on the frontend, you tend to spend a lot of time at the backend,” he said.

But for organisations, such as banks and insurance companies that spent a few years to move some mainframe workloads to distributed architectures while holding on to core banking systems, Kumar said: “What they found was that the processing power they had on the mainframe for the core was just insane.

“We’ve seen examples in the US, where people wanted to go to the distributed side, but the processing power was just not there,” Kumar said. “It goes back to the root cause – when people built software in the earlier days, especially around core banking, they learned to use the hardware capacity of the box whereas modern software today is built to scale horizontally.

“The moment you want vertical scaling, you’ve got to use the power of the architecture that your CPU supports. That’s where challenges start to emerge because of the high volume of transactions,” he added.

Kumar said while there are large banks running mainframe scale workloads on a distributed architecture, doing so requires them to build large powerful servers at scale whose total cost would not vary too much from that of a mainframe.

Rocket Software, which counts more than 25,000 companies – including some of the largest global banks – as clients, offers a platform that lets organisations take advantage of the best of mainframe and distributed computing capabilities.

“We simplify the integration between the mainframe and the distributed side by giving you a technology that does not need you to learn Cobol or work on a mainframe, and yet lets you move data back and forth to run transactions on a mainframe and use a distributed code base on the frontend,” Kumar said.

He added that the company also touts capabilities to help organisations build restful application programming interfaces (APIs) using a low-code and no-code interface, which can then publish the APIs to connect to business processes or core datasets on a mainframe.

“You don’t need Cobol skills to build these APIs, which sit outside of the mainframe, so it solves that entire skills shortage and the go-to-market issue,” Kumar said. “But it doesn’t mean you don’t need any Cobol skills at all, because you still have to do some development on the mainframe.”

Still, Kumar said having those APIs will alleviate a significant chunk of work undertaken by mainframe teams to handle integration requests, because of the sheer number of new apps and interfaces that are being built all the time.

Even as more organisations are building new applications using cloud-native architectures, it is unlikely that mainframes will go away anytime soon, as those applications would still need to connect to a core banking system.

“You should build all modern apps on a modern platform and that’s what we are propagating, but every time you do a transaction it hits the core banking system,” Kumar said.

“I was talking to the infrastructure team of a Singapore bank who said that, historically, the maximum workload on the mainframe used to be at the end of the month when companies were processing payrolls, or the beginning of the month, when there were a lot of ATM withdrawals.

“Today, it’s between 12-1pm every day on weekdays, because of peer-to-peer banking on the distributed side, hitting the mainframe for every transaction,” he said.

Privately held Rocket Software acquired ASG Technologies in April 2021 to bolster its infrastructure management offerings. In 2022, they started going to market together and integrated their product teams, growing revenues by about 20-30%. Kumar said the company expects to grow by about 150% this year.