For the last 20 years we are hearing that banks are going to get rid of their mainframe footprint.
Mainframe is like a white elephant. The running cost for mainframe is extremely high.
And it poses several problems for the enterprise:
- Mainframes manages transactions in a robust way but to make it useful customer engagement needs workaround and integration of mainframe data with digital channels
- Many times the modernization efforts for mainframes end up solving one problem and introduce other problems. For example – there has been a trend to automatically transform mainframe code to .Net functions and objects but that led to .net application that is several times harder to maintain by usual .Net developers and has usually been associated with significant number of bugs.
- Mainframe application development takes longer cycle time in terms of development, deployment, testing and it is not easy to adopt DevOps practices for mainframes
And the changes to mainframes are hard to make. Many Mainframe developers have re-skilled themselves to new technologies and have moved on.
The availability of developers for mainframe systems is as low as any other-end-of-life technology. So now that’s a problem.
So, do we get rid of Mainframes or we do keep them?
Getting rid of Mainframes is expensive. Keeping and supporting them is also riddled with lot of challenges.
What do you do? Whatever you want to do, keep the legacy footprint as a part of your overall architectural landscape. Do invest time in understanding the challenges of legacy systems of the enterprise, how you can deal with them.
It’s not about having a perfect solution; it’s about being practical and not trying to boil the ocean in the enterprise.
Since legacies will stay with traditional enterprises (large banks, telecom companies, Insurance customers), invest in learning how to integrate with these legacy systems so that data from these core systems flow to the digital ecosystem when required.
And CIO’s will love an Architect who leverages the decade old investment.