We've been talking recently about how IBM sees itself just as much a hardware company as a software and services group -- and in particular, about the alleged resurgence of the mainframe. Since IBM is almost the only purveyor of said technology and the vendors of competitive but market-trailing offerings have not beaten a path to our door, we have only IBM's word for it.
Except that now there's an analysts' report out, buttressing IBM's case, which Big Blue is happily touting to press outlets. But does the report in fact support IBM's case for the mainframe?
The rationale for mainframes makes sense on the face of it. Enterprises are focusing on lowering costs of ownership, and in particular, the costs of powering the humungous numbers of servers they've been accumulating over recent years.
Typically, each server runs only one application. That approach brings benefits of isolation between applications, so that if one box falls over, the rest hum away happily. But in today's world of robust virtualisation, super-powerful servers, rising energy costs and heat generation, combined with limits on the amount of power many enterprises and data centre managers can shovel into their facilities, this approach increasingly doesn't hold water.
And indeed, all research, from analysts, from vendors and from IT managers, shows a clear, one-way trend: companies are consolidating, moving their servers into one place and increasing utilisation rates by stuffing them with memory and tasking them with running a number of virtual machines. The servers themselves becoming larger and more capable too but each is likely to be running up a dozen or more VMs, each with at least one application -- although some will have several -- contained within it.
So the movement away from single-application physical servers is clear. Yet the report, from analysts at Illuminata, compares the cost of running small servers against consolidating the applications they're running into a mainframe. Its conclusions are that mainframes cost less to run in admin costs, software licensing, and power, and are cheaper to buy than a whole bunch of equivalent one-off servers.
It said: "IBM demonstrates that mainframe TCO advantages are 30 to 60% better than 30 Sun servers or 300 Linux servers." Yet in today's world, that is a given.
As we pointed out in our news story on the launch of the report, the idea that one server runs one application is no longer the norm in today's data centres where virtualisation means multiple applications contained in each physical server. With the growth of market leader VMware and others such as Xen and Solaris, the numbers of virtual machines and server utilisation levels are shooting up as a result.
We wrote about one such consolidation exercise, conducted by a major bank, here.
So what's more interesting is whether a mainframe is more cost-effective than a handful of big-iron servers containing a bunch of consolidated VMs.
Unsurprisingly however, IBM loves the contents and conclusions of the Illuminata report, entitled 'IBM System z TCO: Man Bites Dog'. IBM mainframe System z general manager Jim Stallings said: "The IBM System z mainframe offers a superior economic proposition for businesses with large or mixed-workload environments. The IBM System z's innovative architecture allows for breakthrough levels of TCO to meet the needs of business today."
However, by lighting on this report and comparing itself favourably to Red Hat, IBM is doing itself and its customers a disservice. Apart from making the wrong comparisons, which its customers and potential customers are savvy enough to see through, it runs the risk of trashing its own, highly lucrative server business. You can almost hear the steam billowing from the ears of IBM's blade and Unix server managers' ears from here.
Why should we care? To a large extent, that's an internal IBM problem, one might argue.
Yet IBM presents itself as a credible source of information and exerts -- due to its longevity, size and market presence -- significant influence on the industry. If Big Blue is to retain that position, it would do well to think -- as IBM founder Thomas J. Watson was wont to urge -- before jumping on the latest piece of apparently good news with such alacrity. And that would be the worst outcome for IBM too.