Like several thousand others, I made the trek to Orlando, FL, in February to attend the SHARE conference. In addition to the z10 mainframe announcement, which provided the overshadowing news, I observed some other things that stuck with me.
First, the z10. The next evolution of mainframes in the IBM line, the z10, took no one by surprise. Strangely, IBM had held its launch of this platform in New York, not at SHARE—perhaps for logistical reasons such as proximity to headquarters or Wall Street. My understanding, however, was that the box shown at the announcement was an empty tin shell; to see the unit in its glory, SHARE was the place to be.
The z10 is remarkable for many reasons. Following the lead in the IBM announcement, many journalists characterized it as a “super server,” along the lines of HP’s Superdome, but at a fraction of the cost and without the aging architecture. This isn’t correct, though I understand how non-tech pundits might have misunderstood. Superdome is a PA-RISC processor-based box, scalable to 64 processors, running HP-UX and HP server virtualization technology. By contrast, the z10 is, well, a mainframe. It has a mainframe operating system and sports mainframe—not server—architecture. Benchmarking places it well above the performance of Superdome (or anything else I’ve seen in the clustered x86 market) and its new (and old) IBM software functionality—born over 30 years— dwarfs anything available in the x86 server world.
IBM might have added to the confusion by contextualizing z10 as a platform capable of consolidating upward of 1,500 x86 servers. They also made hay over the fact that a z10 was about the same size as a couple of traditional open systems server racks and consumed less electricity.
However, contextualizing mainframe value in this simplistic way doesn’t do justice to the differences between mainframes and so-called “open systems” servers. It allows press folk, most of whom have never set foot inside a mainframe data center, to make comparisons between two different technologies as though they’re on a level playing field.
One article published after the show noted that Microsoft’s Mainframe Migration Alliance was tracking 85 projects well under way within companies intended to move apps off Big Iron and on to x86 platforms running the Microsoft OS. The piece went on to say that another 55 projects were started this year. Of course, there’s no information provided about the projects in question or the software being migrated—and perhaps some of these apps could be hosted more cost-effectively on x86. But the upshot of the piece was that mainframes were just bigger servers that could be readily unplugged and replaced by less pricey x86 boxes.
The other favorite pundit jab was to look at the announcement within the context of the VMware virtualization craze. Again, IBM invited this comparison by spending a lot of time in its announcement comparing VMware solutions to the much superior mainframe virtualization technologies, based on LPAR and Sysplex. Forget the $3,000 per virtual server host license and go with Big Blue for server virtualization at $600 per virtual machine. Again, there was truth in the IBM claim, but it also prompted some writers to see the mainframe as just another server virtualization platform.
Moreover, the net result was to give VMware another opportunity to make unsubstantiated claims about its inevitable ascendency in the server virtualization world. Their paid analysts were quoted everywhere to say that VMware was going to raise server resource utilization on par with mainframes: Right now, server utilization efficiency is about 17 to 20 percent, while mainframe efficiency is 75 percent or higher. In other words, VMware stole IBM’s thunder and used Big Blue’s own marketing against them. The nail in the coffin was the announcement, timed with z10, that VMware was about to become part of the standard shipping kit on servers from leading x86 platforms such as HP and Dell.
Bottom line: Now that SHARE is over, IBM may do well to spend a little time stressing the differences, rather than only the commonalities, between mainframes and servers. Using an analogy, VMware servers may be the iPods of present day computing, but mainframes are the dream audio/video systems with more watts, better clarity, and greater value than anything you can buy with micro-size headphones.
IBM had better start helping people understand the differences if they’re going to realize the bump in mainframe popularity they’re seeking. I noticed that SHARE had fewer attendees this year than in the past, and a very small percentage were younger folk. Enough said. Z