Mainframe is dead. Long live mainframe.

Data centre mainframe
(Image credit: Bigstock)

From the early 1990s onwards, the mainframe has been portrayed as the dinosaur of computing, being steadily overtaken by more glamorous platforms and newer technologies. UNIX, LINUX, Windows, distributed systems, networked computing, and grid technology have all emerged to claim their place at the heart of corporate IT. IBM’s System/360 and Boeing’s 747 have been ranked as the two most successful commercial ventures of the mid- to late 20th Century. Both revolutionised their own industries; affordable long-distance travel became possible for the many instead of just the few. Similarly, many organisations were able to automate many clerical processes.

While the number of mainframe deployments may now be in gentle decline, the existing users still have very strong ties to the mainframe; some are seeing up to 40 percent annual growth in processing power. When reviewing the rapid growth in all forms of data in his installation, one UK CIO recently said: “even if my mainframe data declines to become only one percent of my total, this one percent will be the data that runs my bank.”

Many of the uses of mainframes date back to the 1970s and sometimes earlier. The investment in skills, applications, and database formats to date is therefore considerable and cannot be easily cast aside. Many attempts have been made to show other platforms to be more economic but, taken in the round, the mainframe still compares favourably.

IBM came to prominence in the 1960s with its proprietary S/360 architecture. At its peak in the mid-80s, the IBM mainframe dominated around 80 percent of the commercial IT market world-wide. The cost, proprietary nature and inflexibility of this platform eventually created a demand for cheaper, more flexible alternatives, satisfied first by UNIX, then Wintel and now LINUX. Although IBM’s overall market share may have declined since the mid-80s, the mainframe is still highly attractive to this company because no other vendor provides the relevant hardware or software.

Particularly in finance, banking and insurance environments, long-established IT departments with many years’ experience still have corporate databases held on mainframes. The cost and complexity of converting applications from IMS or CICS or DB2 formats to open systems have allowed IBM to maintain a very lucrative position in enterprise computing. Similarly, the increasing shortage of skilled mainframe staff has reduced the ability for installations to change; those who learned their mainframe skills thirty years ago are now close to retirement.

Although proprietary and lacking elegant GUIs, the mainframe platform still sets the benchmark for reliability, availability and security; typically, a mainframe user will see 99.999 percent availability, a significant advantage over many UNIX and Wintel systems. Unbeatable availability, predictable performance, and a proven track record in security and data integrity still look very attractive today, 45 years after the first S/360 mainframe was launched; the resilience and systems management inherent in this architecture still lead the industry and its security standards have also evolved to levels beyond the current capabilities of alternative platforms.

Although today the overall number of sites or organisations using mainframe computing is unlikely to grow because of consolidations, outsourcing or acquisitions, their individual sizes and importance are increasing. The mainframe market can therefore be visualised as a pyramid that is shrinking its base while growing taller; fewer but larger sites, with heavy and continued dependence on this mature platform. Also, by keeping key applications carefully away from open systems environments, the mainframe community already enjoys critical data protection through high security and resilience. On the other hand, open systems vendors and users are struggling to match these levels of service. As a result, today mainframe users are always quick to describe their data as “the crown jewels” of their IT organisation; very few are struggling with unplanned growth, extended downtime, and lost or corrupted data because most storage and data management challenges lie elsewhere.

Although for the mainframe storage administrator overall data growth is slower, more controlled and predictable, data backup and recovery in the mainframe world present similar challenges to those encountered in distributed computing. Heavy batch loads create significant and complex backup activity; full backups are simpler to manage than incremental ones, they are slower and more cumbersome, but on the other hand incremental backups are quicker, but recovery from them is much slower. This forces the storage administrator to trade off the speed of backup against the speed of recovery and therefore a weekly mix of the two is often used. Mainframe data is rarely changed and is typically less volatile than data stored on other platforms, so repeated full volume backups can needlessly contain many copies of unchanged data.

While it is undeniably true that today IBM is the sole provider of mainframe servers and OS, there are many alternative sources in the marketplace for applications and management tools. A quick Google search shows a thriving global community of developers wanting to add value to the mainframe with a plethora of offerings.

The real issue currently facing mainframe users is not the proprietary nature of the platform, but the breadth and depth of the pool of available skilled resources to manage this environment in the next decade. Over the next five years, many long-serving mainframe specialists will be reaching retirement, so the real challenge for IBM will be to develop the relevant systems management interfaces in such a way to overcome the looming skills shortage. Greater automation and simpler and more intuitive GUIs are urgently needed if IBM wants to protect its no doubt considerable revenue streams coming from this market.

Unless the growing issue of the declining pool of mainframe administrators is successfully addressed, further outsourcing is the likely outcome. This would be bad news for IBM, bad news for the assorted companies providing software and services to the mainframe market, but most of all, bad news for the users of the 10,000 or so mainframes deployed across the world today. These organisations would be forced into either a platform change, an expensive application rewrite, or down the outsourcing route, with a potential loss of control of their key business assets. While there is not an absolutely defined date for fixing the large and growing challenge of the mainframe skills shortage, the next three or four years will determine the future of the mainframe. Watch this space.