Why you can't move a mainframe with a cloud
Time to get hybrid
Comment Will cloud computing finally wear the mainframe market down like thousands of years of harsh weather?
Anyone selling a server or a cloud against a mainframe will surely say "yes" to that question. Cloud computing is another style of utility computing, but it clusters cheapo servers together instead of using a large monolithic machine. But the people who actually buy and use mainframes are not so down on the big iron. And that's because they have not just studied history, but lived it.
Successive generations of new computing architectures have ground down the mainframe market from its mighty heights in the 1970s and 1980s, but the venerable big bad boxes are still humming away out there, cranking through untold billions of transactions and making IBM many billions of dollars of revenue and a large chunk of its profits each year.
Both BMC Software and CA Technologies derive a large portion of their sales from mainframe-related products, so they take the pulse of the mainframe market from time to time just to make sure they still have a business. With IBM beginning its ramp of the System zEnterprise 196 servers in September - the machines were announced in July - all of the players in the mainframe racket are hoping to have quite a few quarters of increasing sales as the estimated 6,000 to 7,000 mainframe shops in the world go through their next upgrade cycle.
Yup, that's all there are left. And, generally speaking, they have much larger machines and lots of MIPS crammed into their big iron chassis. And because this new wave of Unix and x64 boxes that is crashing on the mainframe is highly virtualized (like the mainframe has been for decades), IBM and Unisys have to do something to keep the boxes relevant. That's why both IBM and Unisys have created low-cost specialty engines (sometimes using the mainframe engines, sometimes using outboard servers) to accelerate workloads running natively on the mainframes.
(In the case of Unisys, the company has gone so far as to port its MCP and OS 2200 operating systems to Intel Xeon processors and put in an emulation layer so code can move, unchanged, from its ClearPath CMOS mainframe boxes to ones using Xeons. IBM has not done this as yet).
For a lot of complicated reasons that are great news for IBM and Unisys, companies are planning to stick with their mainframes, even as they expect to adopt cloudy infrastructure that, for all intents and purposes is what VMware president and CEO Paul Maritz has been calling "the software mainframe, the mainframe of the 21st century" for the past year and a half.
But even as x64 servers are getting more mainframe-like, the mainframe itself is adapting to this world of virtual servers, virtual storage, and virtual networking. That is what this System z Blade Extension (zBE) running Power and x64 blades in the same management domain and in a hybrid fashion - what IBM calls a "system of systems" - is all about. IBM is trying to make the mainframe the command and control center of pools of Unix, Linux, and maybe one day Windows systems that have been virtualized.
This message does not seem to be resonating with mainframe shops yet, but the IBM marketing machine is only getting started. Those with mainframe expertise will want to believe what Big Blue is saying about the "system of systems" as they try to keep themselves relevant, just like IBM and Unisys themselves in the modern data center. If you can't beat them, maybe you can control them.
In the fifth annual survey of the mainframe base done by BMC Software, here, the software company queried to 1,707 mainframe shops to get their views on the mainframe (by which BMC and CA both mean the IBM mainframe) and its place in the present and in the future. Side note: with that many shops being queried against a base of 6,000 to 7,000 customers, the BMC survey could actually be a representative sampling - something we don't see very often in IT surveys.
(You can have Bill Miller, president of BMC's mainframe service management business, walk you through the survey results here in this video).
Mainframe shops seem to be a little less optimistic about mainframes in 2010 than they were over the past two years, with 57 per cent of those surveyed saying that they believe the mainframe platform will grow and attract new workloads. But a smaller percentage (37 per cent) agree that it will be a viable platform, mostly running legacy applications and not attracting new workloads. Only 4 percent of the mainframe shops polled by BMC said that the mainframe is not a viable platform and companies should consider their exit strategies over the next five years.
Despite the wobbly economy and the desire to reduce IT costs cited as the largest factor affecting mainframe shops, 84 per cent of the companies polled by BMC said they would keep their MIPS steady or grow them in the coming year. The 16 percent that said they were cutting MIPS is a smaller share of the base than back in 2006 and 2007, when the global economy was doing better but when mainframe MIPS capacity was more expensive and zIIP and zAAP specialty engines (which boost the performance of DB2 databases or Java and XML workloads, respectively).
The funny thing about specialty engines, according to the BMC survey, is that they are a large-shop phenomenon, with a relative handful of shops having lots of them and somewhere around a half saying they have none and a quarter to a third (depending on the engine type) saying they had one or two. To be more precise, for example, 5 per cent of the mainframe shops that BMC talked to said they had 17 or more IFL engines for running Linux, but 50 per cent said they had none and 29 per cent said they had one or two IFLs.
About 63 per cent of these shops did not have zAAPs and 20 per cent had one or two, with 3 per cent saying that had 17 or more. The zIIPs have a slightly different distribution, with 47 per cent saying they have none, 32 per cent saying that they have one or two, and only 3 per cent saying they have 17 or more zIIPs activated on their machines. In each case, as the IFL, zAAP, and zIIP count goes up, the site count goes down.
This is about what you would expect among large shops that are trying to consolidate Linux workloads onto the mainframe (or avoid very high z/OS licensing fees by coding a new application to run on Linux instead of z/OS) and to cut costs further on DB2, Java, and XML workloads. IBM is clearly hoping to repeat this trick with the zBX blade extension and the Power and x64 blades for creating a hybrid system, and if it does indeed work - and IBM gets a suitable midrange zEnterprise machine into the field perhaps next year - the zBX could have a lot more success (as measured by customer penetration, not necessarily revenues and profits) than the specialty engines did.
To be fair: if IBM didn't cut mainframe software prices as well as offering specialty engine offload, the mainframe business would have crashed years ago instead of hitting a new equilibrium between an estimated $2.5bn and $3.5bn a year. And, by the way, the advent of the zBX hybrid computing approach has not diminished the appeal of the specialty engines, with more than half of the shops surveyed saying they would be adding more such engines in the next 24 months.
BMC did not ask mainframe shops directly about their plans for deploying zBX blades as bolt-ons for their mainframes, but roughly two thirds of the companies that BMC talked to said that having a unified operations management toolset that spanned z/OS, Linux, and any possible future operating systems (such as AIX on Power blades and Linux or Windows on x64 blades) was important or critically important.
And just for fun, 10 per cent of the mainframe shops that BMC talked to said that using their System z machines to run cloud computing or SaaS applications (the definition was a bit loose there) was an important priority for them in the coming year.
Clouds in my coffee
CA Technologies did its own survey, and rather than just whispering the C-Word in one part of the survey, it based the whole premise of the conversation with 300 mainframe shops on this question: Is the mainframe the ultimate cloud platform?
This study, which was conducted only in Europe by Vanson Bourne on behalf of CA Technologies and which you can read here, had 79 per cent of mainframe shops saying that their mainframes would be "an integral part of their cloud computing strategies. Of those polled, 70 per cent said that cloud computing would sustain or extend their mainframe environments, 74 per cent said the mainframe will have a role in any cloud computing initiative at their company. Survey respondents came from the United Kingdom, France, Germany, Italy, Benelux, Scandinavia, Russia, Poland, Turkey, and the Czech Republic.
The mainframe has its issues, as any proprietary platform does. But the IBM mainframe is under stress in particular because of the graying of the workforce that understands how to make these machines sit up and bark. Averaged across all countries, about 44 per cent of the respondents to the CA survey said the aging of the mainframe workforce and the difficulty in finding replacements for retiring workers will make the mainframe less viable in the future. Still, 82 per cent of the companies polled in the CA survey said they expected to use the mainframe the same or more in the future as they do today. And the big reason why? Because mainframes are reliable and understood, even if they are pricey.
All of this talk of cloud computing and hybrid architectures probably sounds familiar to people who have been around data centers for a couple of decades. Minicomputers surrounded mainframes two decades ago and sucked workloads off the big iron, to be followed in short order by Unix machines and then client/server workloads. Then along came the commercialization of the Internet and its Web technologies. And somehow, thanks to the vast fortunes that mainframe shops have spent on databases and applications, the mainframes have persisted at thousands of companies.
The fact that x64-based private clouds can arguably be built with mainframe-like capabilities will no doubt pull even more applications off some mainframes. But IBM's zBX strategy could, if it has some real meat to it instead of some pretty block diagrams, pull just as many workloads back onto a hybrid mainframe that gives mainframers more control and IBM more profits until the next wave hits. We'll see how this tug of war turns out in about two or three years. ®