Wednesday, September 1, 2010

IBM's cloud grab: the next generation of the mainframe monopoly

In July, the European Commission launched its antitrust probes of IBM's conduct in the mainframe market only four days after IBM's presentation of its new mainframe generation, the zEnterprise.

The Commission took its decision four months after French open source startup TurboHercules had lodged its complaint. That time span is the Commission's goal under its best practice guidelines.

So the EU didn't mean to spoil the party, but there is an important factual connection between the two events: the zEnterprise is an overtly aggressive move by IBM to leverage, expand and extend its mainframe monopoly with a view to enterprise cloud computing. It is high time for intervention to avert abuse that would otherwise cause irreversible damage to the emerging cloud computing market.

In its report on the zEnterprise launch, V3.co.uk picked just the right headline:

"IBM zEnterprise mainframe
assimilates Unix and Linux servers
Brings Power 7 and [x86] servers under its control"

The word "assimilates" alludes to Star Trek, but it is appropriate. IBM calls the zEnterprise a "system of systems" and a "datacenter in a box". Indeed, the objective is to absorb x86 (Intel and compatible) servers. "One box to rule them all", one might say.

If this only meant more competition on the x86 side and if customers really had alternative options, I would welcome it. However, the way IBM leverages its mainframe monopoly is abusive and anticompetitive. Let me explain.

Enterprise cloud computing: clarifying the term

Wikipedia defines cloud computing as "Internet-based computing, whereby shared resources, software, and information are provided to computers and other devices on demand".

There are different ways in which enterprises can operate cloud-based services. They may operate their own cloud (meaning they have the infrastructure in house) or use services provided by third parties. In many cases they'll do a combination of both.

Another distinction is that those cloud-based services may be available only to a company's employees (private cloud, like an "Intranet"), to select business partners (public cloud, like an "Extranet"), to the general public, or a combination (public-private cloud).

Mainframe-managed data: the lifeblood of large enterprise clouds

In every one of those setups, the mainframe legacy comes into play. In most cases -- especially in the most important cases -- new enterprise cloud services are not stand-alone creations totally detached from other business operations.

Other business data (and the applications managing them) are the lifeblood of enterprise clouds. For an example, whatever an airline might do in cloud computing, it will usually have to be connected to the reservation system and/or operations management system. Whatever a bank does in the cloud, it will usually need access to account management and the related transaction processing. Whatever an insurance company does in cloud terms, it will usually need access to all of the essential records.

Where do those essential data (and the applications managing them) reside? In most cases, on mainframes. In this recent blog posting I already mentioned that 80% of the world's business data reside on mainframes. That's the percentage across all industries. In the ones I particularly mentioned -- banks, insurance companies, travel reservations -- the number is even closer to 100%.

So to make an enterprise cloud fly, it still has to be tethered to a mainframe in many cases. That metaphor may sound paradoxical. It's just the sad reality of a lock-in of enormous proportions.

A senior IBM executive noted that IT projects don't start on a "green field" these days:

"We ought to look at these things the way we look at a city -- a city is a living, breathing thing, and you don't literally rebuild New York every year: you add to it. And more often than not, you're renovating what's already there -- you're improving what already exists, you're not replacing what exists."

IBM's all-absorbing zEnterprise cloud machine

The zEnterprise (latest mainframe generation) was designed from the ground up to connect mainframe legacy workloads with new cloud computing technologies.

It doesn't really add much new on the original mainframe (System z) side. CPU clock speed went up only from slightly below to slightly above 5 GHz, and the number of processors from 64 to 80 (in each case, 16 are reserved for internal purposes). But the key new element is that the zEnterprise is a "system of systems", integrating x86 (meaning Intel or Intel-compatible) blades to a greater extent.

I've seen and used the integration of different computer architectures in the same system a long time ago. The Commodore 128 had a CP/M component with a separate CPU, and my first PC was a plug-in card for the Amiga. That was fascinating, but those devices were toys.

What IBM now wants to achieve with the zEnterprise is that companies consolidate their entire data centers on the basis of IBM's technology, putting tons of IBM's x86 components under the control of a mainframe. IBM calls it a "data center in a box". I consider it a very dangerous expansion and extension of the mainframe monopoly. Let me explain what's wrong with this.

The need for integration

There are technical reasons for which it does make sense to run a mainframe legacy application on the same system -- not just in the same network -- as new cloud applications that require access to those data and the applications managing them.

One very important aspect is administration. If you run a large IT operation, you need an efficient way to keep track of all the systems, all the time. IBM's Tivoli systems management software can be used to administrate both System z (mainframe legacy) and x86 resources on a zEnterprise. Tivoli is proprietary software and there's no indication that competing vendors of Intel-based hardware would have access to its interfaces.

The need to integrate mainframe legacy workloads and new enterprise cloud applications distorts competition. For the mainframe, IBM has a monopoly. Consequently, Big Blue also has a monopoly for a "system of systems" including the mainframe. This expands, extends and exacerbates the monopoly. The original monopoly (mainframe) is leveraged to create an even broader cloud-related monopoly. It's like one monopoly "spawning" another.

IBM denies its customers an important choice

Customers should have the choice between two different paths to a "datacenter in a box":

  1. the zEnterprise path: bringing x86 hardware under the control of a mainframe

  2. the virtualization path: executing mainframe legacy workloads on Intel-based servers

The second choice is the one IBM denies its customers, and it's a very important one. It would allow many customers to make their purchasing decisions based on new cloud computing needs and to achieve similar performance at a potentially much lower cost.

IBM allows customers only to put the old cart (the mainframe legacy) before the new horse (cloud computing). As long as customer don't have an alternative to the zEnterprise approach, they are going to be overcharged and the lock-in that already exists today will only exacerbate in the future, resulting in ever-increasing costs and less innovation.

There are no technical reasons for not offering the second choice. The Hercules open source mainframe emulator is a reliable and innovative solution. It's a mature piece of software whose development started in 1999, and today's Intel-based hardware is powerful enough for many legacy workloads. The only problem is that IBM doesn't allow customers to run the proprietary z/OS operating system (which is key to execute legacy workloads) in emulation. That restriction must come to an end.

Even those who decide otherwise would benefit from the second choice because it would put pressure on IBM and result in more competitive pricing.

Regulatory intervention can open up the market, restore competition and safeguard innovation. In order to do so, it must be timely and decisive. Now is the time.

If you'd like to be updated on patent issues affecting free software and open source (including but not limited to the antitrust investigations against IBM), please subscribe to my RSS feed (in the right-hand column) and/or follow me on Twitter @FOSSpatents.