Q&A : HP storage and server CTO UK & Ireland, David Chalmers

The mainframe is dead, get ready for the flashing blade

Chalmers: "The mainframe's replacement will be large scalable blade server farms."

Mainframe platforms have been the mainstay of firms with large transactional and/or data processing requirements since the 1960's. Over the past five years there has been much speculation that such systems would eventually have to be replaced. The major reasons being the high legacy software support costs.

Additionally, many of the IT personnel overseeing the hardware and software pieces of the platform will be close to retirement.

Computing talked to HP's UK & Ireland storage and server chief technology officer David Chalmers about what might replace these systems, and how he thinks firms currently reliant on mainframes should proceed.

Computing: Are mainframe systems living on borrowed time?

David Chalmers: I think they are, there are two problems, one is the cost of the software. IBM talks a great deal about the hardware costs, but nobody cares, because the monthly mainframe software charges that people pay are so high.

But the second and more fundamental reason why I think firms will move away from mainframes concerns the people who deal with the software used by such systems. Those are the people who know, run, develop, maintain and put new functionality into the applications and business solutions that run on mainframes. They're in their late 50’s and 60’s, and are retiring.

Would firms having these employees upgrade their terms of employment to keep them there for longer?

You certainly see some of that going on, but these people are the tail-end of the 'golden pension' generation, and most have final salary pensions. I think they're the last of the generation who can afford to say – “nah, I don’t need to do that.” But even if they do get financial 'upgrades', that just jack up their costs.

How are firms migrating away from mainframe hardware?

What they’re doing is taking pieces of workload away from the mainframe as quickly as they can, but they're doing it in waves, rather than one big hit. It's rarely a wholesale 'take everything off in one lump and shut it down,’ process. It tends to be moved as a result of business logic first, and then application logic. Data migration tends to follow afterwards.

Won't re-writing my business critical Cobol-coded applications be expensive?

Cobol can be ported to other platforms and there are some vendors, MicroFocus for example, who specialise in doing exactly that. However, the more widely used model incrementally removes application logic to newer, more effective platforms as new functionality or processes are developed.

So what actually will replace the mainframe in these scenarios?

Firms are looking for a platform that’s at least as reliable, and performs as well as the systems they’re looking to replace. It's got to be more flexible as well.

A mainframe is still at heart a big single system, capable of running many different workloads. A modern converged infrastructure can do all this, but also takes scale-down to the next level. Most companies looking to replace their mainframes are looking at much larger, scalable blade server farms.

The replacement will need to run many different workloads, as well as many different applications and operating systems all run dependent on need. When demand is low, not only can the system take down workloads that are not needed but it can power down elements of the physical infrastructure as well, saving power and emissions.

Does Oracle's acquisition of Sun make a new competitor for you in this market?

No - is the honest answer. Oracle is one of HP’s biggest partners, and there are more Oracle databases run on HP than other vendor systems.

Larry Ellison has said that Oracle's business model emulates that perfected by IBM back in the 1960’s in that it involves completely integrated systems. That’s one of the scariest comments for customers that he could make, because customers are very uncomfortable with it, it implies 'lock-in'.

Are Oracle really going to take the hardware market seriously? I would question that, especially if you look at the cancellation of the high-end Rock processor project. I'd also question where the billions and billions of dollars for the R&D for the processors is going to come from.

What do CIOs see as their biggest challenge currently?

Some of the CIOs I talk to still have the same old challenges. Namely, how do I innovate, drive my business forward, but spend less money doing it. It's been like that for 30 years. I think it's the shape of the answers to those challenges that is changing.

Cloud computing, as it becomes a reality, rather than some nice ideas and architectural drawings on PowerPoint slides, will help. However, the big challenge with this technology is going to be governance.

If you're going to start adopting cloud technology, the first thing you have to decide is, what services am I going to implement myself and what services could I buy from the outside.

I was talking to a big oil company recently, and they said they had between 50 and 70 external service providers delivering technology into their organisation. How do you manage that, and how do you integrate the external services with the crucially important ones you've kept inside their organisation?