The future of mainframes in the enterprise

Mainframes tend to divide opinion amongst IT professionals. On one side there’s the virtualisation-driven view that such machines are relics of a bygone technology era and should be replaced or migrated to virtual, cloud servers as soon as possible. This is already happening in areas such as SDDC (software-defined datacentres) that are increasingly taking on roles traditionally held by mainframes.

On the other side are the advocates and engineers who believe that these big, robust computers still have a role to play in today’s enterprise. Who’s right?

If the engineers on the front-line are any guide, there’s little growth in demand for mainframe services. Newsgroups dedicated to mainframe engineers, such as IBM-MAIN@LISTSERV.UA.EDU (registration required), are peppered with requests for work referrals in amongst the technical Q&As. It appears to be getting harder for people in this line of business to find new roles, at least in some locations.

However, market analysis doesn’t necessarily support this anecdotal evidence. A modest growth rate over the next five years is predicted. In that period, the general consensus seems to be that more sales will be at the higher-performance end, rather than at the entry level where Windows/Linux servers are increasingly competitive.

It’s worth bearing in mind that sales of new machines don’t tell us how many organisations are still using the old ones. The market may not be growing by much, if at all, but there’s life in it yet. There’s also a belligerent confidence on the part of those still involved, coupled with a large dose of cynicism about potential replacements.

Dave Wade, a mainframe engineer since 1976 who recently worked on migration of physical servers to VMware environments, is bullish about mainframes. “I don’t think anyone techy involved in mainframes can see any point in moving off them,” he says. “Managers think they can save money by getting rid of mainframes because PCs are cheap. In reality, delivering scalable enterprise software on any platform is challenging and expensive as it needs new design, capacity planning, etc.”

Al Kossow, curator at the Computer History Museum, told me, “Whatever you replace [a mainframe] with will be obsolete in 10 years, and won’t be supported as well because that’s the way the economy works now.”

When individual mainframes are retired, it’s for any one of a number of reasons, including companies going out of business or enterprise-wide replacement of core IT systems. But it’s the migration of mainframe operations to virtualised systems that’s currently in favour. Security fear is one factor driving this, with no shortage of analysis and comment (try a search for “mainframe security”) pointing out the risks of using mainframe systems. Another issue is the retirement of the people who know how to program these systems properly, though judging from the newsgroups there are still plenty of skilled engineers available.

Intuitively it makes sense to migrate from an unwieldy chunk of metal to a scalable, secure, virtualised platform. However, just because a concept is intuitively appealing doesn’t make it logically or commercially sound. “Vendors want to push you into cloud subscriptions because that helps with vendor lock-in. Companies like it because they can outsource their hardware and eliminate support people,” says Kossow. “[But they] then have people twiddling their thumbs (or lose business) when the network goes down.”

In addition, the security fears may be overblown. An IT manager for a major APAC banking organisation, who understandably wishes to remain anonymous, told me that the bank’s support time and costs for mainframes are dwarfed by those of newer systems, especially when it comes to security. “We had some dumb policies in place, based on the idea that a secure perimeter meant our internal systems were invulnerable to malware. When I finally managed to force through a proper audit, we found that every single workstation was infected. No such problems with the mainframes, which plodded on regardless just doing their job.”

To make matters more complex, any transition process is fraught with risk. Al Kossow again: “The two biggest problems with migration and retirement are that companies count on a few people to know their business process and how that was implemented by computer, and feature-creep/requirements change in the system design and replacement processes.”

Wade is more blunt about migration problems. “As a manager, do a simple risk analysis as to how likely it is you can get enough code moved to get your bonus and get out before it falls down in a nasty heap because the mainframe guys have shafted you and/or you didn’t put enough resources in to build the scalable platform,” he says. “Then do a Google search for failed mainframe migrations and re-calculate the risk…”

There seems to be a generational divide when it comes to mainframe technology. In researching this article I contacted the director of the Vintage Computer Federation, Evan Koblentz. He told me, “Younger employees assume anything from before their generation must be obsolete.” He went on to liken mainframes to Cobol: “Bunch of non-technical technology reporters kept writing in faux shock that important organisations ‘still’ use Cobol. Damn it, those companies use Cobol because it’s good not because they are stupid.” He believes the same applies to mainframes, and it’s clear that he’s not alone.

There may even be a temporary growth in mainframe popularity underway. As Koblentz says, “mainframes are having a resurgence because of big data: regular servers can’t keep up.” This may be a short-lived blip rather than sustained growth. “The resurgence may end in a few years if NMVe-based servers become common,” Koblentz tells me.

Even so, big iron obviously still has a role to play in today’s enterprise IT environment. To misquote Mark Twain, reports of mainframes’ death appear greatly exaggerated.