What Might be Possible for the Future of the Mainframe
While we don’t know what IBM has on its roadmap for the future of the mainframe, we can look at what has happened and try to draw inferences. The whole exercise is a hostage to fortune—just as anyone who has ever published an article saying that mainframes will die out any day soon like the dinosaurs did.
What We Do Know
The first thing we can hopefully agree on is that the majority of organizations using mainframes are unlikely to be seduced to other platforms. They are running mission-critical applications on CICS or IMS and they those applications won’t work anywhere near as well on an alternative platform. The truth of the matter is that there are some companies whose level of business is falling below the threshold that makes mainframe computing viable. They are the organizations that will be featured in computing magazines as migrating off old-fashioned mainframe technology. In fact, a mainframe today is as different from a mainframe from 50 years ago as a Formula One car is from its predecessor from 1966.
The next thing we know for certain is that computing power increases all the time and the size of components gets smaller—see Moore’s Law. Does this mean that mainframes will one day fit in a shoe box? The answer is no. A mainframe has to look the part, i.e., bigger and more dominating than distributed computing. Any space saved by reducing the size of one component means that space can be filled up with other components or more of the same component. Over time, components also become cooler to run and can be placed closer together. That makes future mainframes even more powerful than they are today.
What Needs to Change
Future hardware needs to be flexible. Through the use of firmware, PR/SM-like capabilities or VM, it might be possible to virtually reconfigure all the components to work in such a way that as many VMs as you want with as much virtual hardware as you like will be easy to create and change on the fly. That means IBM needs to keep hold of its hardware division so that it controls the innovations built into its mainframes and also to ensure that there are innovations that can be built into the hardware.
Something has to change with device addressing. People have been predicting the end of IPv4 for almost as long as they’ve been predicting the demise of the mainframe and yet both are still with us. IPv4 uses a 32-bit address which allows for just over 4 billion addresses to exist. Clearly, there are more than that in existence. Various techniques are in common use to get around the limitation. IPv6 has been the obvious answer for a number of years. IPv6 uses 128-bit addresses, which means it has 3.4×10^38 addresses. With the growth of the Internet of Things (IoT), more addresses are needed every day. Either IPv6 will reach critical mass and everyone will just say, “Oh alright” and start adopting it, or some alternative addressing method will have to be devised that will become the must-have method. Either way, IPv4 will, by necessity, become a thing of the past.
It won’t just be ATMs that connect to mainframes (and usually IMS), it will be any number of devices. You may not use a mainframe to control your home environment, but if you were a big bank, you may very well use the sensors connected via the IoT to control the temperature around your vault. In these financially worrying times, you might use a sensor telling you how many people come into the bank to determine whether there was a run on the bank and to stop anyone else from entering.
One area that will see interesting software developments will be the Liberty server. I think we can safely predict that just about any software will be able to talk to CICS, IMS or DB2. Those stalwart backend megaliths will be able to connect to browsers (which they can already) and any of those IoT devices (including drones with sensors and cameras on them). Application development will be faster because parts of existing code will be available for reuse. This will be possible through repositories of available APIs that programmers can make use of.
I also predict that more GUI-like software will predominate. Younger users expect it, it’s available and the only reason it hasn’t happened yet is because older users can do things quickly using the familiar green-screen technology. As these veterans hand over the reins, the new-look interfaces will predominate. I predict the rise of Artificial Intelligence on the mainframe. Everyone else has AI and I’m sure mainframe AI can take over a lot of the running of the mainframe—especially during the period when the experts have retired and the newcomers are getting up to speed. After all, IBM already has Watson, it’s just a case of porting it to a mainframe. And I guess there will be a lot more cloud everywhere.
The Future Is Still Unclear
I assume that the names will change. I know some people who still talk about MVS. But the marketing people will, one day, decide that z has had its day. Maybe the new name will be an emoji? Perhaps they’ll change to a symbol (after all, there aren’t any letters after z in the alphabet)?
It might be that IBM is hoping to make Linux the dominant OS. You’ll run Linux on your laptop (you already do on your phone or tablet if it’s Android) and you’ll run it on your mainframe. You’ll have wonderfully transferable skills. Perhaps z/OS will run under Linux, or perhaps CICS, DB2 and IMS will be ported to Linux. It would make running big data easy because it already runs under Linux.
This isn’t a look into the deep future and I have no idea what IBM is really planning to do. I just thought that these ideas were possible and based on current trends. What do you think?
Trevor Eddolls is CEO at iTech-Ed Ltd, an IT consultancy. A popular speaker and blogger, he currently chairs the Virtual IMS and Virtual CICS user groups. He’s editorial director for the Arcati Mainframe Yearbook, and has been an IBM Champion every year since 2009.