Why there's life left in the mainframe yet

The conventional wisdom says that the days of the mainframe are numbered, but it just might be that a new generation of apps will breathe new life into it.

A state-of-the-mainframe survey by Syncsort shows that the mainframe market as ever is experiencing some shrinkage, but also that there are some grounds for optimism. Syncsort has been working on mainframes since 1968 and has customers in more than 85 countries. ZDNet spoke to Syncsort's chief product officer, David Hodgson.

There were two sides to the survey. One was to have a general understanding of where the users thought the mainframe's place was in their current thinking, and the second was to get a better understanding of how they were using their mainframes for analytics.

According to the survey, as far as mainframe people are concerned, they're not going anywhere. Over 70 percent saw it as critical to their business for the large-scale transaction processing that they were doing.

The lesson is that for those people who are still using mainframes, they are critical to their business.

We do see a few cancellations. With such a big penetration of our Sort product, we have quite a good feel of the market the whole time and some of the smaller customers are cancelling. That's mainly because they have moved off the mainframe.

Right now, it's got down to the very smallest VSE shops. I think all the cancellations are in VSE. And then you see a smattering -- 19 percent -- saying that they are going to get off in two to five years, and 10 percent saying six to ten years.

But talking to customers on a regular basis, they say that these days it is hard to get even a three-year plan together because things change so quickly.

These days the accent is on flexibility so how does this change things?

This comes under the heading of, what are people doing to modernise their back-end processing?

I think there are two factors here. If you look at the old model where people had their systems of reference and systems of engagement, you probably don't want a lot of change in systems of reference. You do want to be actively changing and adapting to new needs in your systems of engagement.

What that means is that with your systems of reference you want to be flexible enough that you have the systems to get at the data and get at the transaction processing. You have to be flexible enough around your systems of engagement and I think people have cracked that nut.

Even around CICS systems, there are a number of ways for you to put APIs on the front of those. Once you have done that, then you can do your mobile apps and your people won't even be aware that it is a mainframe on the back-end. The important thing is the API, and with this whole move to wrap things in APIs nowadays -- and remember you can run Java in z/OS now -- these new systems won't even be aware that it's a mainframe running the backend.

What were the other key findings in the survey?

We were looking to probe any movement off the mainframe and the governance around that. How easy is it for people to do that? And we wanted to look at this new emerging world of Log Analytics.

The sort of questions we asked were general questions, not necessarily directly addressing the mainframe.

We asked what the main corporate objectives were, and the number one answer was meeting security and compliance requirements. Meeting service level agreements was number two, and number three was reducing CPU usage and costs. That was a big shift.

Reducing costs has been the number one requirement for a long time and to see it drop to number three was interesting. I think that pointed to the way that meeting service levels is increasingly a problem. I think the driver behind that has been the need to improve access and, I think, the driver behind that is most likely mobile devices.

We asked customers, if they were doing analytics what platform would you do it on?

Overwhelmingly if they are going to do analytics they would move the data off the mainframe. That is what customers are doing now and I would predict they are going to go on doing that. The skills for new things like machine learning aren't on the mainframe.

Do you think the mainframe is declining or is it holding its own?

If you talk to people they will tell you that the mainframe budgets are definitely under pressure. People see them as flat.

Now we are not a public company and we don't disclose things that we don't have to, but I can tell you that our traditional mainframe revenue is growing.

There is no indication that mainframes are disappearing. What that means in detail is this -- I divide our customers into three groups:

1. Those at the top who really see the value of the mainframe and are happy to continue investing in it.
2. Those who think that if they could replace the mainframe they probably would. But then they have mission critical systems and the know that they have nothing that would scale the same way.
3. Those who are maybe getting towards the end of a long migration off the mainframe.

But we have been talking about this issue for so long that we know that those top two groups are, by far, the majority of the population.

The only thing is that one of the persistent problems for the mainframe is the skills problem. When you think about it, the pioneers of IT were the baby boomers and what did they grow up on -- the mainframe.

Obviously, the baby boomers are all retiring and this is a problem for the industry and the mainframe and one of the issues is the lack of skills.

But then there are quite a few youngsters who understand that the mainframe is not going to go away completely anytime soon, so, they think, maybe it is a good place to start building a career.

The other issue, I think, is that this whole area of the move towards machine learning is one of the things that will help. I believe that as we move towards more sophisticate systems with machine learning, any lack of mainframe skills becomes less and less of a problem.

Read more on mainframes

• IBM Machine Learning brings Spark to the mainframe
• IBM adds machine learning knowhow to its mainframes
• Syncsort buys mainframe expertise to link Big Iron to Big Data