Commercial Interests and the Future of Linux

Linux seems to be central to IBM's future plans. IBM claims to have invested billions of dollars in Linux. IBM is not a non-profit company it surely sees revenue opportunities in Linux that are not present in Windows. However, IBM has never clearly specified why it prefers Linux to Windows.

As a technology Linux is not radical. It does not offer anything noticeably different from what Microsoft is offering. For whatever reason IBM is investing in Linux it clearly is not any technical superiority over Windows.

The issue of price can also be ruled out. IBM is working with Redhat and SuSE, companies that are interested in generating revenue from the sale of Linux. Redhat licensing fees are substantial but this has not bothered IBM at all. If IBM wanted to avoid the licensing fees it could have done so by purchasing a major Linux distribution. Novell paid 210 million dollars for SuSE, this kind of money is pocket change for IBM.

IBM is not into Linux because of technical superiority, it is not motivated by pricing differences, and it does not wish to control the distribution of the OS either. So why is IBM getting into Linux?

IBM is not the only company that likes Linux. SGI has placed an even bigger bet on Linux. SGI initially tried to enter the NT workstation market to attain better volumes and reduce hardware and software development costs. Another motive was to increase the pool of software available for SGI workstations. SGI's NT foray turned out to be a complete disaster and worsened the company's financial situation.

What SGI learned from its NT experience was that it is not possible for a technology company like SGI with low volumes to compete in the Wintel market. Wintel is ruled by companies with no research and development budgets but high volumes and extensive experience managing the supply chain. SGI realized it cannot leverage its technological edge in the Wintel world. The biggest hurdle in bringing proprietary technology to Wintel is Microsoft's control of Windows.

Microsoft centrally manages the design of its operating system. Microsoft's OS designers look at the market and decide what new features they want to add to the OS. Customers do get a say in this process but ultimately Microsoft implements features required by the vast majority and rejects changes helpful to only a few customers. Microsoft also cooperates with Intel and chipset manufacturers to ensure any features it adds to the OS will be well supported. This results in a level playing field with very little to differentiate hardware vendors from one another.

The Microsoft model is very well suited to companies like Dell. Dell is a fabulous company when it comes to manufacturing boxes on the cheap. Its supply chain management skills are unsurpassed. However, Dell is a dinosaur when it comes to technology. Dell depends on Microsoft and Intel to do all the research and development and come up with designs it can manufacture cheaply and sell at a profit.

Such a situation is not helpful at all to a company like SGI with a strong technology portfolio. To effectively compete with Dell, SGI needs features in the OS that can be leveraged by its hardware. Take the example of the XFS filesystem. XFS is a core SGI technology. It is a filesystem aimed at applications needing high throughput IO. XFS is essential to asserting SGI's leadership in storage related solutions.

SGI would have liked XFS integrated into Windows. Microsoft on the other hand has no benefit in integrating XFS into Windows. A filesystem although more or less a modular piece of code can still require changes to the kernel and regular maintenance. Microsoft sees no positive return on investment on non-mainstream features. Another issue for Microsoft is of code ownership. Windows is proprietary software and Microsoft can not allow all and sundry to submit code into Windows.

In contrast to Windows, Linux is very open and non-centralized. Companies willing to invest programmers can contribute features they require. Linus is happy allowing new features in as long as they don't degrade the kernel for the mainstream user and add some value. SGI realized this and was able to bring XFS to Linux by contributing programmers to work on Linux.

IBM's motives are similar as well. IBM wants to reduce hardware and software development costs and leverage its technology leadership. IBM also realizes that in the future Linux will enjoy much wider software support than any proprietary Unix. Windows is bad for IBM for precisely the same reasons it was bad for SGI. IBM has much better volume than SGI but still it is unable to compete with the likes of Dell.

Dell is a player in the Linux server market as well but SGI and IBM are no longer competing with it. If someone wants a low-end Linux solution they can go with Dell but a high-end solution is only provided by IBM or SGI.

IBM has another advantage, it offers a complete line of Linux solutions. Dell might be cheaper in terms of hardware prices but it does not offer complete solutions. Some pieces of hardware have to be purchased from other vendors. Integrating and maintaining hardware and software from many vendors can be an expensive proposition, training and maintenance costs are higher. Consequently, total cost of ownership can be higher. The Dell price advantage gets negated by Dell's inability to offer a complete product line.

By going with Linux, any company with strong technology can enjoy access to cheap hardware components and a bigger market without being handicapped. Such a move simultaneously reduces OS development costs. Technology companies like Linux as it lets them sell hardware and is good for the bottom-line.

The very reasons that are so attractive to technology companies also explain these companies' lack of interest in the Linux desktop. Desktop computing is about off-the-shelf hardware, a standard OS, and low prices. IBM could push Linux on the desktop and it might even manage to sell some hardware but ultimately IBM will be expanding Dell's market. When the market becomes big enough Dell will enter and takeover. For the time being Linux adoption will be limited to former proprietary Unix vendors.

Strong Linux adoption does not mean proprietary Unixes will disappear. Linux is good enough for many things but not for everything. The XFS implementation on Irix offers many features not supported on Linux. Features such as GRIO (guaranteed rate IO) require realtime functionality. AIX too has advantages and IBM is not giving up AIX. AIX will be retained as a high-end solution.

Maintaining a proprietary OS along with a Linux strategy seems contradictory to the original goal of saving on software development costs. How can a company save on software development costs if instead of maintaining one OS it has to maintain one and a quarter OSs? To achieve cost savings companies will cut down development on proprietary OSs to focus only on the features Linux does not offer. Proprietary OSs will increasingly become domain specific and have kernels attuned to performing specialized tasks.

Linux kernel forks will occur but only to add radical functionality and will not be intended for the mainstream. Forking mainstream Linux is not viable. Maintaining a domain specific OS and a mainstream OS are two very different propositions. Just like the chip fabs the cost of maintaining a mainstream OS doubles every few years. The codebase of such an OS grows very rapidly and the maintenance costs increase along with it. Unix forked but that was in the past when OS complexity was low. As OS complexity increased most of the Unix forks died. The remaining Unixes are dying like flies.

SGI uses a custom Linux kernel for its Altix systems. Technically it is a fork but SGI goes to great lengths to de-emphasize that. SGI wants to see most of the customizations merged back into Linux. This will decrease future maintenance costs. The last thing SGI needs is having to maintain a Linux variant. Smart companies will avoid forking Linux. The not so smart companies will quickly go out of business.

SUN does not understand the advantages of Linux as well as IBM and SGI. SUN is vulnerable and knows it but continues to stick to a Solaris centered vision. Solaris maintenance is getting quite expensive for SUN. If maintaining Solaris was cheap then why did SUN give up on Solaris x86? SUN has backtracked on Solaris x86 but likely because AMD is offering it the deal of a lifetime on Opterons. The smart thing for SUN is to embrace Linux the same way IBM and SGI have done. Solaris will stay as the high-end offering but will become much more interoperable with Linux.

In the future continued contributions from companies like IBM and SGI will turn Linux into a first class platform for high performance computing. Linux will support very diverse hardware suitable for very many uses. Windows will continue to run only on mainstream hardware and this will become a liability. Every large corporation needs some fancy hardware. Linux will integrate much better with the fancy hardware while Windows will not. Overall Linux will become an excellent choice for technical and high performance computing.

by Usman Latif  [Dec 13, 2003]


OSNews discussion for 'Commercial Interests and the Future of Linux'
Scaling Linux to New Heights: the SGI Altix 3000 System
Kernel Fork For Big Iron?
IBM to spend $1 billion on Linux in 2001
Novell buys SuSE Linux