This article is part of an Essential Guide, our editor-selected collection of our best articles, videos and other content on this topic. Explore more in this guide:
2. - Hardware: Server refresh options: Read more in this section
- Introducing the new norm for server processors
- Traditional vs. appliance servers
- Open source server design takes off
- Finding the right server memory
- New mainframe technology validates big iron
Explore other sections in this guide:
- 1. - Is it time for those servers to go?
- 3. - Server software upgrades: The other side of the refresh
The mainframe is not going into terminal decline. On the contrary, mainframe technology is evolving to keep itself relevant and necessary to enterprise IT operations.
IT observers have declared impending doom for the mainframe for at least the past quarter-century. They bBlame the proprietary architecture, the lack of developers, the aging-out of mainframe administrators and big iron's inability to match scale-out PC server farms and grids. Taken together, these old arguments should havekilled the mainframe by now, but with mainframe development on the right track, none of them holds true.
This is an era in which IT seems content to spend the same amount each year on computing -- focusing much of its efforts on saving money. In 2012, after a 10% to 20 % dip in revenue in the first three quarters, IBM saw mainframe revenue jump 56% in the fourth quarter. IBM's mainframe business is on a three-year streak of revenue growth. Its Unix/Linux (System p) and Windows (System x) alternatives saw declines in 2012. In fact, if 2012 was proclaiming anything, it was the eventual death of Unix/Linux. But I don't believe that's true either.
The mainframe fully reinvented
About five years ago, I suggested that IBM make the mainframe a "hub" -- a fully networked node tuned for certain types of workloads, so that data centers don't have to choose between the mainframe and something else and can't have both. IBM has focused mainframe development on turning big iron into a full participant in the enterprise architecture. All of IBM's major enterprise software products -- administration, security, data management, development -- are on approximately the same track inside and outside of the mainframe, and most apps can easily and even dynamically move between mainframe and non-mainframe servers. This makes it far easier for data centers to use the mainframe flexibly, taking advantage of its inherent robustness, security and high-end transactional scale-up.
IBM's mainframe development plans for 2013 appropriately are less dramatic than they were in the past five years. The "bridging" architectures of zEnterprise and PureSystems, IBM's emerging converged infrastructure line, form an adequate foundation for future mainframe development, including enterprise architectures.
Data center administrators and IT developers will use future mainframes as more or less transparent parts of a fully modern, overall enterprise architecture. The mainframe still fails to support Windows in public cloud architectures adequately, which could limit, though not reverse, mainframe use.
Few argue today that the mainframe's proprietary architecture cannot keep pace with the evolution of hardware and software technology. IBM says that a new generation of system administrators has adequate skills to operate mainframes; some businesses are buying new mainframes with a preference for z/OS up front rather than Linux. There are few major barriers to software, either.
Evolving mainframe technology is attracting new customers, even in the U.S., because the old knocks on the mainframe no longer apply.
Mainframes' offerings for big data, scale-up computing
We all hear periodic outbursts of "the death of …." Sometimes this death is justified -- no one who had the dubious joy of programming for dumb terminals mourns their death. But some architectures are proclaimed dead while they're still very much alive, stultifying innovation on them and wasting their unique strengths for next-generation computing. For very good reasons, the mainframe refuses to stay decently dead.
The mainframe pushes the limits of scale-up computing. While scale-out grid systems have achieved some significant successes -- Google made notable advances in language comprehension and translation with its thousands of servers' divide-and-conquer approach -- on a per-processor basis, mainframe scale-up's tight integration offers performance advantages over non-mainframe scale-out's loosely coupled networking. Mainframes also offer advantages for total cost of ownership, robustness, security and run-the-business-app experience.
Wayne Kernochan is president of Infostructure Associates, an affiliate of Valley View Ventures. This document is the result of Infostructure Associates-sponsored research. Infostructure Associates believes that its findings are objective and represent the best analysis available at the time of publication.