Mainframe Decommissioning Isn’t a Modern Approach Anymore

"I predict that the last mainframe will be unplugged on March 15, 1996" -Stewart Alsop Jr in 1991

Yet here we are nearly 30 years later, and the mainframe is alive and well. The people who wanted to see the mainframe die in the 80s and 90s are old at this point or getting there, their opinions and goals no longer make sense in the 21st century, which is ironic because mainframe critics will often call mainframers old and out of touch.

The world of those who want the mainframe to die was an entirely different one than from the one we live in now, back then things were moving at a much quicker pace than today and there was an obsession with new technology far more than today. While yes there's still been crazy improvements in the 21st century I would argue they're more evolutionary than revolutionary, all of the fundamentals of our current tech were pioneered in the past century when you think about it. Our workstations and servers have gotten smaller and sleeker but at the end of the day they have most of the same characteristics as they did 20-30 years ago. What we need now is a more pragmatic stance rather than the trend chasing of the 20th century, despite the best efforts of mainframe critics "the great die off" of the mainframe failed miserably and now more than ever before we need the advantages of the mainframe to tackle major issues facing the world.

Interest in mainframes is slowly ramping up among generation z and for good reasons. We're in a horrible climate crisis and our dependence on data processing and storage are contributing to it further, x86 has failed to address these issues, meanwhile the mainframe is more efficient than ever before and absolutely crushing the competition in terms of reducing power consumption and having high efficiency.

With the ability to run Linux on IBM mainframes and ever expanding z/OS modernization initiatives like Zowe and z/OSMF, mainframes have become far more user friendly than they were back in the previous century, which is eroding another one of the few real advantages competing platforms have had over mainframes historically.

Many organizations that run mainframes have ended up with a messy landscape of servers that include x86 and IBM mainframes, IT leaders that wanted to see the mainframe die built tools off of the mainframe and migrated small bits off of the mainframe. Because the sheer scale of mainframe applications tend to be so huge in the institutions that run mainframes they often end up still having a mainframe, this is incredibly inefficient as you end up having to have more people with little or no overlapping skills and need to maintain vastly different ecosystems. It makes far more sense to centralize as much as possible on your mainframe, this is especially feasible if a lot of the x86 applications run on Linux. Centralizing your applications onto the mainframe when feasible would be far more financially and environmentally sound, rather than dispersing them over a weird mix of mainframe and x86, I would argue migrating onto the mainframe is the truly modern approach of those two.

An alleged advantage that is often thrown around when people talk about mainframe modernization that involves migrating off of the mainframe is the cloud's ability to scale better, that is mostly nonsense. IBM Z just like the cloud can be scaled up in a variety of different ways, such as buying more processors for the system in a similar fashion to the cloud and having multiple mainframes work together in a parallel sysplex. Scaling vertically like the mainframe does can also make application development more simple, scaling horizontally can be very messy and inefficient for some workloads.

The "skills shortage" is often mentioned as well, as a young person myself I don't believe it's an issue in the same way mainframe critics do, it's real in the sense that yes we do need more mainframers but it's entirely a self-made issue. Companies have been neglecting the mainframe for decades, since most schools have stopped teaching mainframe skills, companies will have to actually do the training themselves through training programs and apprenticeships.

I hope the rest of my generation sees the value of the mainframe if we ever get to run things, because I think a lot of current IT leaders are going in the wrong direction.

Bonus: if you want to see how little the fundamentals have changed I would recommend watching the video from bell labs on UNIX.