March 4, 2009 The case for opening up the mainframe market
By Roger Bowler, mainframe professional and creator of “Hercules”
I have been following the legal battles between IBM and Platform Solutions Inc. (PSI) and T3 Technologies (T3) over the last couple of years with great interest. As the founder of the Hercules open source mainframe emulator project I feel that we are impacted by many of the same issues that put both PSI and T3 out of business. As a mainframe IT professional, it bothers me that there is no longer any competition in the mainframe platform space. While all the press coverage related to these cases has helped to raise awareness of how important this is, the fundamental issues sometimes get lost. In reality, these are very clear cases and once people understand the seriousness of the situation I hope action will be taken to open up this important market. In my mind there are three major issues here:
- The IBM mainframe is a unique platform that is incredibly important for enterprise computing.
- Through its actions, IBM has prevented alternative mainframe solutions from being viable and as a result it now controls 100% of the mainframe market.
- With no competition in the mainframe platform market, prices have remained high and customers have fewer choices than if there were more vendors creating and selling alternative solutions.
Let me provide some more detail on each of these areas.
The Mainframe is Unique
While some might argue that the IBM mainframe is just a big server, based on my 20 years working with mainframes and mainframe professionals, it is clear that mainframes are in a class of their own. Mainframe applications use unique languages, systems and interfaces that simply don’t exist together on other platforms. Mainframe systems use very different concepts, terminology and processes than any other platform. Mainframe professionals who write applications or administer mainframe systems have unique skills that take years to master and are different than the skills required for any other platform. IBM states that mainframes have their own “style of computing.” I couldn’t agree more. Simply put, mainframe workloads must run on IBM-compatible mainframes and cannot simply be moved to other non-mainframe systems. Ask any mainframe professional and I think they’ll agree with me on this.
This level of uniqueness doesn’t exist for any other major business platform in the industry. Java applications can be run on almost any platform. UNIX or Linux applications can easily be moved to virtually any hardware vendor’s platforms. Windows applications can run on servers from any of the leading server manufacturers. Yet a mainframe application can only run on an IBM mainframe with IBM system software. If the mainframe market were dying or if mainframes were no longer used in mission critical environments, then maybe this wouldn’t be a big deal. But based on news from IBM, mainframe use is increasing and mainframes continue to be a central platform for the world’s biggest businesses and governments. IBM recently reported that the shipment of System Z computing power grew by 12% over the year. IBM also claimed recently that IBM mainframes “power the top 50 banks worldwide and 22 of the top 25 US retailers.” Some analysts estimate that a quarter of IBM’s US$100B in revenue comes from the mainframe. Clearly the use of mainframes is not only important to both IBM and its customers, but the business is also growing.
IBM Controls the Mainframe
Not that long ago, there was a flourishing market for IBM plug-compatible mainframes. In fact, companies such as Amdahl, Hitachi, Olivetti, Comparex, RCA, Fujitsu and others competed with IBM in the mainframe hardware market for over 20 years. Over the last decade, a new wave of solutions came to market that used a new approach to emulate mainframes in software. As a European, it was great to see many of these innovations coming from European companies such as Comparex, UMX, Transitive and others.
I was the original author of one of these mainframe software emulators – the open source project called Hercules. By creating a virtual machine that allowed mainframe system software and applications to run on commodity hardware, people who wanted their own mainframe for development or testing now had the option to run a virtual mainframe on their PC. A group of ex-Amdahl engineers created their own emulation technology at PSI and Fundamental Software Inc. (FSI) became successful selling its FLEX-ES mainframe emulation product. Both of these technologies were designed to offer customers the ability run production mainframe workloads on substantially lower-cost hardware.
However, all these technologies – both the plug-compatible mainframes from the last century and the new software-based emulators – needed cooperation from IBM in order to be viable. IBM mainframe system software is needed to run IBM mainframe applications. For decades, IBM provided licenses to the customers of both the alternative mainframes and the solution providers that had platforms based on software emulators including FSI, T3, UMX and PSI. As the cost of engineering compatible hardware-based mainframes grew, the plug-compatible manufacturers gradually exited the market. However, the emulator vendors were able to use innovative software to keep pace with IBM’s new offerings and it appeared that there would be a flourishing market of alternative technologies for hosting mainframe workloads.
Several years ago, I was optimistic that this flurry of innovation in the mainframe emulation space would help to reinvigorate the mainframe market and make it more viable for mainframe professionals like myself. However, IBM made the decision a couple of years ago to discontinue all licensing of mainframe software for use on non-IBM systems. This has effectively killed all the competitors in the mainframe market and left us with only IBM as a provider of mainframe platform solutions.
The goal with the Hercules project has always been to provide the thousands of mainframe devotees around the world with the ability to run mainframe software on their PC but currently there is no legal way to run z/OS and other current IBM system software on Hercules. This is a real shame. The situation today is that IBM is the only choice for mainframe platforms, despite the fact that technologies exist today that would allow customers to run mainframe workloads on alternative platforms. By only allowing IBM mainframe software to run on IBM mainframe hardware, IBM has been able to eliminate all mainframe alternatives and totally own the mainframe market.
The Mainframe Market Needs to Be Open
While I am not an economist or a marketing expert, I am a mainframe professional and in my day job I develop software for mainframe customers. With no other legal alternatives in the market, software developers and other mainframe users are at the whim of IBM’s pricing and licensing policies. Following the abrupt withdrawal of FLEX-ES from the market, many independent software vendors needed to find a new platform for developing mainframe solutions. Our only choices were to buy a new expensive IBM mainframe or lease time on one of IBM’s mainframes. Yet, as the creator of Hercules I know that we could be using much cheaper systems for doing our development. It is too bad that IBM will not allow independent software vendors to use Hercules or FLEX-ES or PSI’s technology to develop new IBM mainframe applications. There are rumors that IBM may offer a mainframe emulation package for developers in the future but this would still mean IBM is the sole provider of mainframe platform technology. Why not open up the market to a broader ecosystem of solution providers? In addition, some of our customers would love the option to run some of their mainframe workloads on a much smaller mainframe, such as the ones that T3 used to sell. Yet again, with IBM as the only option in the market, we only have IBM’s very expensive offerings as an option for running mainframe workloads. This not only impacts us but these higher costs must ultimately be passed to end users and customers. This just doesn’t seem right. Not right for mainframe users and developers and not right for consumers who pay higher prices for services hosted on mainframes.
Summary
I strongly believe that there needs to be fair and open competition in the mainframe market. Customers should have the ability to choose platforms that meet their needs not the needs of a single vendor. I believe Hercules users and others should be able to legally run mainframe applications on non-IBM platforms. I’m not asking IBM to give away its software or intellectual property. I simply want mainframe users to be able to license IBM’s software on reasonable terms for use on non-IBM mainframes. If this were to happen, I actually believe that the mainframe market would expand and IBM would benefit through greater adoption of mainframe technologies. With fair competition in the mainframe space, mainframe users and consumers would also benefit as the cost of mainframe computing would likely come down to prices closer to distributed servers. Opening up the mainframe market would be a win-win for all and have huge benefits for customers who rely on mainframe applications.
Note: We welcome your comments on this article using the form below. To send an email to Roger Bowler, please use our contact form.
Reader Comments (7)
The economists and anti-trust regulators go crazy making this a complex subject. Congratulations, Roger, for describing the situation in real-world terms.
Roger, I am a big fan of Hercules and have been following the decline of competition in the mainframe space for some time. Thank you for presenting such a clear and concise case for why we need to allow Hercules and other solutions flourish in this very important market.
By driving PSI and T3 out of business, IBM has put mainframe technology on the slope of inexorable decline.
No new business will paint itself into a corner by putting any of its key business processes on an IBM mainframe. Now I know full well that all sorts of new businesses paint themselves into the exact same corner with Visual Studio, but the cost of entry is generally under $10,000 while the mainframe cost of entry takes a big chunk out of a million bucks.
By making mainframe technology unviable for new businesses, IBM has restricted mainframe technology to those businesses that are stuck with / committed to mainframe technology. In the current economy, some of these businesses are failing. Over the next several decades mainframe usage will decline because of the inevitable attrition of a static captive customer base.
IBM mainframe technology would face better future prospects if new businesses saw economic benefits in using mainframe technology, but that would require attractive applications at entry costs comparable to Windows and Linux.
Well stated and well put. A clear and concise description of the history and the current state of mainframe computing.
What has been missing ever since the drop-out of the Amdahls, Fujitsus etc., the very sudden disavowal of hercules (see the Redbook-alteration issue), the PSI Lawsuit and the FLEX-ES termination, is a statement by IBM as to what is the reasoning behind this.
If it is what everyone assumes, it would be monopolize, expensify and control the market. As long as IBM does not provide any contrary information as to their strategy, this will be the universally assumed motivation.
IBM's failing to voice any comments on the current situation and the continued silence of FSI tells all, doesn't it?
Now please don't get me wrong: Actually all mainframe professionals should profit from pricing being as stable as can be. But while prices remain high, the market is slowly collapsing in terms of size. I severly doubt that IBM's figures of a growing mainframe market are reflected in a growing NUMBER of mainframe users - instead I fell that solely the AMOUNT of mainframe utilisation is growing, without increasing revenues for the ancillary mainframe service providers, especially ISV's. This collapse results in a slow but unescapable diminuation of the entire mainframe market, be that for software or service providers beyond the chosen few big ones.
So I concur with Roger, the current state of affairs more or less rings the bell for the "last round": There are many that say IBM is milking the last big revenues out of users struggling to get away from the mainframe and not really watching out for what comes after. I might be wrong, though, but there is not much to be heard from big blue apart from marketing bla and small to very small users are more or less lost when trying to run a big iron operating system.
What makes things slightly ludicrous, in a satirical way, is that there are many people within IBM who have voiced (off the record) their concern about this state of affairs and that, for example, their efforts to provide an adequate replacement for FLEX-ES for Developers, although never as good as your own machine, have been tremendous up to now. This part of IBM seem to be trying extremely hard to "offset" the detrimental effects the overall corporate decisions are having on those mainframe users actively supporting mainframes by doing serious development of software. It's a big company, after all, and a major decision will have some collateral damage, I suppose.
Up to this moment, there is no indication of any compromise and thus I intend to add what I have written above to Roger's essay in support of the statements he has made.
God what a bunch of whiners. IBM owns z/os, it's theirs. How they license it is their own business. That they won't let you run it on an a hundred dollar Intel chip and destroy their hardware business is tough on you. Don't like the price then use something else. Go write your own if you want it so badly.
But why does anyone care? Its 2009 for Pete's sake! People still using mainframes are just pouring their money away. Forget them. Mr. Bowler, do what the rest of our generation did, go learn VMS, UNIX, Linux, Apple, whatever, and hate IBM. Want to make a lot of money? Use your expertise to help people migrate away from z/os to an open system.
Guys,
Fully support for open platform. I'm a "youngster" (3 years with mainframes). I think Zed Ohess has fair few on the situation. Here are my points which probably detail a bit Zed's opinion:
- Mainframe - as term - is central computer which acts as central point in IT-infrastracture (i.e it is Main Frame).
- IBM mainframe - is Mainframe technology developed by IBM.
- There are other Mainframe technologies available from other vendors (centrilized computing). HP Superdome is an example.
- There is and always will be demand for high performance computing systems (CPU and I/O) for huge corporations and organizations (this is niche of market). This is how Herman Hollerith made his business.
- I think it doesn't matter for them [organizations] much what it is: IBM, HP, Sun, Fujitsu or Amdahl or something else. It should be reliable, cost-effective, reasonably easy to support/maintain and should meet specified requirements. The decision making, I think, is similar to those of buying new freight truck or plane.
- If IBM Mainframe is unique then all the credits go to IBM - IBM has developed this unique technology, then as owner IBM can claim what it can do whith it.
- IBM Mainframe will be priced according to demand for it. As long there are customers who are ready to pay that big money for that technology, price will remain. If less and less customers will value technology for this money, then more of them will move to other platforms (regardless of what these platforms are; they [platforms] should be as close to IBM mainframe as possible, to minimize migration risks and costs).
- I think most of people (and thus companies) don't like to be boundled to one vendor. There should always be backdoor or plan "B". And those IBM mainframe users who don't like that will (if not already) address that. It's all about spreadsheet-based calculations and checking the bottomline.
- If someone doesn't like IBM mainframe monopolistic approach - get rid of the mainframe. Yes, certanly you will loose some sort of functionality, but that's the deal. Linux wasn't developed overnight. The user-friendly Linux what we see now - contribution of thousands of programmers over the +19 years (if not more) to make open operating system, which gives you functionality (at least basic) of UNIX systems, while also being able to run as desktop operating system. Firefox is another example.
- There are companies on the market which do IBM mainframe-to-other-platform migration. Some of the companies are listed on this website as well.
- As computer professional, for me it doesn't matter much what the vendor is, be it IBM, HP or others. I like high-performance servers and large-scale computing. As long as the Thing computes and technology is good, we can study it and use it on full capacity. It doesn't matter much what language it's either: it takes 6 to 12 month for experienced programmer to get programming-practices and concepts of "new" language. Only Assembler language is (of course) highly platform dependand. But since computers are computers, Assemblers are very similar to each other in concept (registers, base instructions set) and IMHO it shouldn't take much for assembler-programmer to master assembler for another platform, be it IBM mainframe CPU, HP Alpha, Sparc or any other (still, I will stick to 6-12 month, believe it or not).
- IBM technology isn't perfect - the way it's managed and administered requires deep knowledge not only of the technology itself, but the history of the technology as well, as its contains a lot of legacy code working not in the way of current computing practices. For example, there is no filesystem available on z/OS and filenaming is reasonably outdated. We are limited to have files 44 characters long and no more than 8 characters in qualifier. Job names are 8 chars as well. Members of PDS have 8 chars (everyone knows why it'd be great to have more).
- And I think there is a bit of controversy in attempts to advocate IBM mainframe while at the same time trying to say there should be alternative(s). I think such alternative couldn't be IBM mainframe technology itself: it could be worse or better or similar to, but not the same technology (like analog of IBM mainframe).
I'm a big fan of Hercules, and I would personally love to see IBM license its software for use on other platforms. Having said that, There is absolutely no reason that IBM should do that. IBM is a shareholder company, in business to make money. If they were to license the software then they should only do it if they expect to make enough money in license fees to offset lost sales of the hardware. And lost hardware sales drives up the unit cost per mainframe since they will have to spend just as much engineering time to develop them, but will sell fewer units, so they need to make enough money to cover that cost as well. If they were to license the software, the license fees would need to be in the $50,000 to $250,000 range per copy to offset the reduced mainframe sales. And then, instead of people whining that IBM wouldn't license its software, people would be whining about how IBM is overcharging for the software, and IBM might look bad for doing this as well. I think they are best off not to license it.
Instead of wanting something for nothing, I think that if there really was a lucrative market for the software, some company could develop compatible software, and license that to all comers. I don't expect that to happen anytime soon.