No-one who really knows mainframes thinks that they are going away anytime soon. However, that isn’t the commonly held view by people who don’t really understand what a mainframe is and does. In fact, I have written previously about the importance of mainframers explaining the benefits of using a mainframe to non-mainframe specialists in the IT sphere, and management in general.
Notwithstanding the fact that mainframes and the idea of using mainframes has been under attack since the 1990s, it’s useful to understand what mainframes are doing at the moment and where they are planning to go in the near future.
Let’s start with modernizing the mainframe – a term that pigeon-holes the mainframe as being an anachronistic device little changed during its sixty-year lifespan. Firstly, the mainframe is very modern. It has AI integrated onto its chips. It keeps data secure at rest, in flight, and in use. These are things that would have been little more than science fiction in 1964.
So, is using the cloud a good idea for mainframe sites? The answer, like the curate’s egg, is yes and no – using the cloud is good in parts. There are some things that the cloud does better than the mainframe, and I would suggest that applications or parts of applications that need to use those facilities should use the cloud. There are other things that the mainframe does best. Those applications or parts of applications should stay on the mainframe. There seems to be too much all-or-nothing thinking going on in meetings discussing mainframes and the cloud. You really want to cherry pick the best bits of each platform.
The other big topic – in terms of column inches being written about it – is artificial intelligence (AI). Everyone has heard of AI. Lots of people have tried ChatGPT or Gemini (aka Bard). But few people really get what it means. I now get emails telling me that someone is using AI, for example that podcast platforms are now using Ai. And to the untrained eye that looks like a step into the future – but it’s a future built mainly on ideas taken from Sci Fi movies and TV shows. If I have a medical scan and need someone to interpret the blobs on that scan, then an AI is probably exactly what I want to do it accurately. If I want some kind of artificial generalized intelligence (AGI) then I’m probably going to have to wait. And it’s the AGIs that films and TV usually show – usually as the villain!
For mainframers, AIs trained to do specific tasks are great and are worth investigating further. Hoping for something like HAL in the film 2001 is still a long way off. However, using an AI to hack a mainframe is possible because it is a task that an AI can be trained to do successfully. And that means a ransomware as a service (RaaS) AI could be made available to anyone who could pay for it and had a grudge against an organization or who simply wanted to see what would happen. Like every new technology, Ai can be used for good or evil. It certainly makes the idea of using AI to protect the mainframe a very important project to initiate. When I say protect, what I’m really talking about is suspending jobs or users that seem to be doing something untoward, and alerting the security team. The team can investigate further and either resume the job because it is OK, or take further action against the would-be hackers.
Thirdly, with so many mainframers looking to retire in the near future, or who have already retired and are working as consultants a few days a week, there needs to be a way to keep mainframe expertise within an organization. I think there are two main strands to solving this problem. Firstly, mainframe education is needed to help people understand how mainframes work. IBM recently introduced its Mainframe Skills Council as way to close the mainframe generational skills gap. It’s not just IBM, the Skills Council includes Academic Mainframe Consortium, Albany State University, Broadcom, DNB Bank, HOGENT, M&T Bank, Northern Illinois University, Rocket Software, SHARE, and 21CS.
And if enough people can’t be trained to become mainframers, the second approach is to use open-source software (OSS) on the mainframe. Now I know that warning bells sound for some people whenever I mention open-source software and mainframes in the same sentence, however, there is some really good software available that can run on a mainframe, eg Zowe and Ansible.
The reason that an organization would want to use OSS on a mainframe is because there are plenty of people who are familiar with using it on non-mainframe platforms. They can bring their software expertise and use it on the mainframe without needing to know the arcane secrets of the mainframe. In addition, IBM is completely onboard with the idea of using OSS and, in many cases, has plenty of examples of how to use them on a mainframe, making life much easier to use for mainframe newbies with expertise in a particular piece of OSS.
The other thing about mainframes is that they are adaptable and can embrace almost any form of technology that is available on other platforms. Obviously, having Linux as a component of the mainframe helps. Things like blockchain for financial services can be used on a mainframe. And there are plenty of other technologies like that. Things that in April 1964, when the first S/360 mainframe was announced, would have seemed magical or impossible.
And, of course, the AI mentioned above can be used to fill the widening mainframe skills gap. At this stage, with AI, there will always need to be some people around because, as I’m sure we’ve all found at some time or another, there will be circumstances that you haven’t been trained to deal with. That’s where, seemingly clever, AIs currently find their limit. I’m sure in future that this won’t be the case, but in the next five years, it probably will.
A summary of the current state of the mainframe industry is that mainframes are powerful computing platforms that are not only keeping up with the times but are also leading in many ways. They do, however, have a bad reputation amongst some groups of people as being old-fashioned and out of date. This is something that all mainframers need to address, because mainframes aren’t going away any time soon.
No comments:
Post a Comment