Windows and Linux will be the leading growth operating systems (OSs) on x86 by a wide margin through 2014. Solaris will be important, but a distant third in percentage share during the next 24 months, as Linux continues building momentum. The costs of supporting Linux, Windows or Solaris on x86 will be competitively negotiable.

• The Unix segment on reduced instruction set computer (RISC) and Itanium will have a declining market share as applications continue to migrate from RISC/Itanium to x86.
• As the race for market share intensifies among OSs, the OS license and support costs of Linux, Windows and Solaris on x86 will be negotiable and competitively aligned for similar systems.
• With the abstraction of the OS and applications from the hardware, the value proposition is moving up the stack, and the decision point will move with it.

• Create a spreadsheet comparing x86 platform and OS pricing with that of non-x86 platforms. When upgrades or new purchase negotiations arise on non-x86 hardware and software/OS components, aim to shrink the gap (delta), particularly for OS and system software license pricing.
• Set migration plans in increments of 25%, 50% and 75% of the application portfolio during the next five to seven years, because nearly 90% of applications will fit the scaling of x86 servers and blades.
• Pressure Microsoft and other proprietary software vendors to reduce software license and maintenance on their OSs and stacks by setting up a test bench of near-equivalent open-source counterparts (for example, Linux, Open Office, Apache, JBoss) as alternative environments to reduce costs.

The overall server market is expected to see a decline over the period 2008 to 2014, with a compound annual growth rate (CAGR) of -0.4%. However, rather than a constant decline, this will manifest as a sharp drop in 2009, due largely to current economic conditions, followed by a steady recovery. The Linux server segment will be the fastest-growing server OS market, projected by Gartner to have a CAGR of 12.6% (2002 to 2014), but slowing during the next five years (see Table 1). The key reason for the lower 2008 to 2014 growth is the current economic environment. The cost-to-performance ratio, platform openness and increasing availability of applications are key drivers for the aggressive adoption of Linux. More vendors are offering Linux as an alternative to legacy or proprietary OSs (in particular, as Unix replacements) in selling hardware and suite solutions. There is also a trend that more Linux is being used in mission-critical systems but at a slower penetration rate.

The Windows server OS is the only other growth OS with a slower growth rate than Linux, but at a higher annual shipment revenue. The slower growth compared with Linux is reflected in Gartner's assumption of a slower adoption rate of the new Windows 2008 Server OS.

The Unix segment will have a declining market share, with a negative decline of 0.8% during the next six years. Market shares of vendors (HP, IBM, Sun Microsystems) will shrink accordingly. The shrinking market size of Unix is due to the continuous substitution with less-expensive alternatives, such as Windows and Linux, and less-expensive support resources. Sun's growth strategy for Solaris on x86 and its marketing of Solaris as a better OS (compared with Linux) on multicore servers has helped give the x86 Unix category of the total market a positive growth rate.

Server architecture influences a company's OS selection. The server technology can be bifurcated into two main streams — x86 and non-x86 — including Itanium and RISC variants. On a revenue basis, they are within a few billion dollars of each other, but on a unit volume basis, x86 dominates server technology. This makes the selection of the OS a vital and long-term decision. Three primary OSs (Linux, Windows and Solaris) can run on x86 hardware (Linux and Windows on Itanium, but niche applications):

• Linux consists of a variety of alternatives, from Red Hat, Novell, Oracle, Ubuntu, CentOS and many other variants, plus numerous back versions on legacy OSs.
• Windows is in the midst of an upgrade and has been facing stiff challenges from Linux, especially in Web servers that could extend into the cloud.
• Solaris 10 (not to be confused with OpenSolaris) is a freely downloadable OS, and is offered as a subscription service. Sun is late to capitalize on the open-source momentum as it tries to position OpenSolaris as a community-evolving OS (mostly based on Sun-developed OS code).

Within the non-x86 market, there are three leading Unix OSs (HP-UX, IBM AIX and Sun Solaris 10), as well as the IBM z/OS system and others — with most of these vendors competing for IT deployment share. This has created competitive dynamics in which most Unix server vendors are competing against and among themselves, because they all can deliver Windows, Linux and Unix. Independent software vendors (ISVs) — for example, SAP and Oracle — are now turning against Unix, realizing that it's more important to help users reduce their hardware costs and prop up their software licensing. Inevitably, the migration that started in 2000 will continue and will favor Linux and Windows, while Solaris, as the only x86 Unix, attempts to add share.

To understand the OS competition, users need to understand the role of the server architecture and its influence in OS selection. Is the longer life cycle, but more-costly non-x86 architectures the key decision driver, or is the lower cost of x86 hardware but at a shorter operational life more important? The obvious response is to evaluate the applications for growth, scaling, stability, reliability and availability, as well as cost (acquisition and operational), and to determine how many non-x86 applications are candidates for migration to the x86 technology.

An additional factor is also crucial: virtualization. The IT department will have to make virtualization choices about platform, host OS and style (for example, large system hardware and software partitions, clusters, OS host virtualization and hypervisors). Many enterprises that run their current applications in frame-based partitions (Unix or mainframe) may be reluctant to migrate those applications to another architecture (for example, VMware ESX). Consequently, the tightly coupled workloads on high-end Unix (or mainframes, especially when hosting Linux in IFL engines) create a gravitational pull that mitigates against their rehosting to an x86 server (see Figure 1).

ISV enthusiasm is a major driver behind OS success. This enthusiasm generates demand for products, which, in turn, stimulates more ISVs to increase their focus on the OS. Therefore, a natural cycle is created that maintains the leadership edge of successful OSs, but creates a barrier to entry for others. Some OSs — such as OpenVMS — will only address a niche audience, and may achieve a successful life cycle with a limited set of applications. Added complexity is created by binary compatibility issues among OSs that are alike; Linux, for instance, runs on many platforms, but software compatibility is limited according to architecture.

We've charted the relative size of a software portfolio on the vertical axis, against the enthusiasm and popularity among the ISV community on the horizontal axis (see Figure 2). The shapes signify whether an OS has a diverse, balanced portfolio that suits multiple styles of applications, or where the portfolio is geared toward network-centric workloads (application, Web and infrastructure serving), or data-centric workloads (database management system [DBMS] and vertically scaling applications). It's not necessarily the sheer number of ISVs in the portfolio that's most important, but it's having the right ISVs in the portfolio with a good relationship that is critical to a product's success. This applies to OSs that may not have a large installed base; niche OSs only need the right ISVs to develop for the platform to create a successful product. However, as OSs emerge, whether in the market or on new platforms, distribution organizations will seek to port applications to their OSs to gain market share. Linux is a prime example of this.

New workloads are dramatically more likely to be aimed at x86 platforms (and, therefore, Windows and Linux) than any other OS environment. The only Unix system that compares favorably with Windows and Linux is IBM's AIX, with Sun, HP and IBM's z/OS systems trailing. From these Gartner client polls, although not a scientific sampling, Gartner has concluded that Linux and Windows will be the two primary growth OSs, driven largely by lower acquisition and maintenance costs, compared with high-end systems and the breadth of ISV support drawn by the large installed base of x86 (see Figure 3). Other systems will not disappear due to plenty of new and upgrade hardware Unix contracts, amounting to the million or multimillion dollar range.

Market share will be won on the basis of performance, efficiency, use, agility, availability and security. However, as virtualization brings many of the same features of Unix (such as mobility, high availability and fault tolerance, capacity planning and placement optimization) to x86 at far lower costs, it becomes increasingly evident that the OS is becoming more divorced from the underlying hardware architecture, and more distributed. This will make the OS decision more dependent on external infrastructure variables than just on the features and functions tied to the OS kernel.

IT data centers will face OS consolidation, often at an intersection with Linux. Nonzero-sum market dynamics ensure that Linux momentum will continue. In a nonzero-sum market, competition and collaboration are necessary ingredients for survival. A greater number of opportunities to partner may broaden market opportunities that otherwise might not have existed. Because Linux, in particular, is not a single, one-of-a-kind packaged product, and can be an enabling technology in many markets and vendor products, the opportunities to put down roots are manifold. There are numerous combinations in which vendors may appear to be competing but also that are contributing to the further growth of Linux (see Figure 4).

Examples of nonzero-sum market dynamics are:

• Red Hat vs. Novell: Although both compete, each, in turn, is contributing to the expansion of the market, which is still not saturated.
• Novell vs. Microsoft: Both compete vigorously, yet Microsoft and Novell have an agreement in which Microsoft will help sell Novell certificates for subscription support to ensure that each vendor does not slow the progress of deployments because of intellectual property and interoperability concerns. Each vendor expects to drive business, which otherwise might be slowed due to political and cultural feuding within IT organizations.
• IBM vs. HP: Each competes on its own brands of Unix, but both help drive the Linux market by offering platforms supporting Linux.
• IBM vs. IBM: Strangely, IBM's OS divisions may be competing with non-Linux OSs. But on all of its hardware, IBM has made Linux into an adjunct OS so that enterprises can run Linux on any of IBM's x86 or non-x86 platforms. IBM benefits by offering enterprises a larger portfolio of applications that they would lack without Linux.
• Oracle vs. Red Hat: Both Red Hat and Oracle have been collaborative partners. Despite Oracle's new strategy of supporting Linux (based on Red Hat code), both vendors coexist in expanding the total market for Linux servers. Oracle's contribution to mission-critical Linux can be viewed as expanding the adoption of Linux, whether for Oracle or Red Hat.
• Sun vs. Sun (Solaris 10/OpenSolaris): Sun competes against Linux when it positions Solaris as a Unix OS, especially for SPARC. However, it also is driving the market for open source with its OpenSolaris, MySQL, Java and GlassFish top-to-bottom open-source software (OSS) stack. Users have the option to place these software components on OpenSolaris, Solaris 10 or Linux.

Companies will face as many OS decision difficulties in the x86 market as in the non-x86 server market, although the cost, flexibility and modularity in hardware mitigates some of the issues. The competitive issues revolve around business models, feature/function velocity, physical-to-virtual transition, platform and OS management, and coping with the extraordinary high-volume potential of virtual machines and clouds. Issues that continue to be important are:

• How and when to use open source
• What subscription licensing is appropriate
• How to create or recreate mission-critical infrastructure and business continuity

It all comes down to the question: "How much can an IT organization leverage and harness its current infrastructure as it transitions into new architectures, new licensing terms, multivendor integration and interoperability, vendor accountability, service levels and indemnification?" Overconfidence can lead to higher risk, while too much caution can slow an enterprise from making rapid strides toward IT modernization. When developing an IT modernization strategy, IT planners should become familiar with the key issues and challenges facing the x86 OS vendors, and they should checkmark compliance and strategies on a risk measurement scale (see Figure 5).

As with all OS vendors, Microsoft continues to enhance its products in areas of improved performance, better security, increased manageability, virtualization, and best-in-class reliability and availability. We believe that, for Windows to remain competitive, Microsoft must show that it can continue to rapidly innovate, and gain or retain leadership in as many parts of the market as possible — including high-end online transaction processing, DBMS, business intelligence, analytic processing, scientific and engineering algorithms, real-time applications in financial services, telecommunications and so forth. Despite the challenges, Windows is the market leader for x86 servers and will maintain a leadership position through 2013:

• Scalability: Windows Server 2008 R2, due early in 2010, will support a maximum of 256 logical processors. With eight cores per chip available in the market, the biggest server designs have more processors than can be used in a single, running instance of Windows Server 2008. Because the number of processors will tend to double every two years, the support in the Windows OS must grow at least as fast.
• Hardware evolution support: Hardware is growing not only in size, such as more memory and more processors, but also in the types of technologies used. These shifts require corresponding changes in the OSs to deliver their full potential. Among the changes emerging in the server market are flash memory, 64-bit mode, mandatory hypervisors and more-flexible designs called "fabric-based servers." All these functions need to evolve in future releases of Windows Server in a timely pace to match the wide market adoption of these technologies.
• Manageability: Microsoft must continue to enhance System Center to better manage heterogeneous customer environments, including those with Linux and Unix systems. Part of the endeavor involves tailoring the installed modules of Windows Server to specific roles to move it away from a monolithic, large general image to a sleeker footprint. This approach will simplify the manageability to specific roles that enable data centers to manage Windows to those functions that are needed.
• High-performance computing: High-performance systems in clusters predominantly run on Linux. However, Windows Compute Cluster 2003 and Windows HPC Server 2008 are combining and encouraging an adequate ecosystem that makes Windows-based clusters practical. We don't see Microsoft disrupting the growth and trend in Linux clusters. However, data center managers of clusters that serve a diverse population of users are adding Windows as an alternative mode, particularly for Windows applications that are beginning to scale up.
• Modularity: Microsoft must reshape Windows to be more modular, enabling the modules to be combined in appropriate ways to support many server roles. This approach will enable Microsoft to address additional deployment models, as appropriate — for example, cloud computing and subscription delivery business models.
• Cloud: Microsoft has unveiled its plan to extend Windows Server to a cloud-based, pay-for-usage environment in which customers can develop, deploy and manage applications. This Windows Azure effort will take place in tandem with the ongoing Windows Server development aimed at the traditional licensed, on-premises business and usage model.
• 64-bit applications: The Windows software ecosystem has been slow to port to the 64-bit environment. However, interest in running Windows x64 is increasing, because of 64-bit support in SQL Server and Exchange. Microsoft has declared that the current Windows Server 2008 is the last release that will offer 32-bit mode; thus, all software and drivers for Windows will be 64 bit from Windows Server 2008 R2 and beyond, rapidly driving 64-bit operation into the mainstream. Also, 64-bit Windows can offer dramatic advantages in price/performance for applications supporting 64-bit addressing — notably database serving, terminal serving and Web serving.

We summarize Microsoft's Windows Server 2008 opportunities and challenges in a "pro and con" report card (see Figure 6)

Next, we summarize Gartner's views of the outlook and strategies for Microsoft, Sun and other vendors, including Red Hat, Novell, Oracle, CentOS and Ubuntu. To evaluate their likelihood of success on x86 servers, we have generated three scenarios: most optimistic, least optimistic and most likely.

Strategies for market expansion and growth:

• Exploit new market opportunities in clustering, high-performance computing, cloud and virtualized infrastructures
• Expand portfolio with OSS applications
  • Test, certify and optimize OSS for Windows
• Expand developer loyalties (.NET) and Linux interoperability
  • Test and certify Novell SUSE Linux Enterprise interoperability
  • Use improved interoperability to discourage migrations from Windows
• Reinforce intellectual property through indemnity agreements
• Evolve virtualization with strategic emphasis on comprehensive physical and virtual management

Scenarios:

• Most optimistic: Slows Linux growth through hybrid Windows/OSS
• Least optimistic: Fails to slow OSS/Linux in cloud, licensing model and OS modularity
• Most likely: Retains middle upper tier back-office strength, maintains strong market share leadership, but lacks force in clustered, Web-centric infrastructures

Strategies for market expansion and growth:

• Drive innovative and energy-efficient product design and packaging
• Re-energize the developer community
  • Use OpenSolaris as a better Linux alternative, while exploiting the advantages of OSS
  • Seed loyalties bottom-up through OSS development
  • Shift company to higher growth through OSS appeal in developing economies
• Shift focus and design to high-volume and massively parallel markets and infrastructures using the subscription support model of Linux

Scenarios:

• Most optimistic: Slows Linux, revitalizes Solaris, and generates large pipeline of systems and services
• Least optimistic: Fails to staunch the Linux appetite and ongoing Unix migration
• Most likely: Takes two years to achieve a stable, critical mass in volume-based subscriptions

Strategies (of all named Linux) for market expansion and growth:

• Appear unified, while competing for share growth and migrations
• Drive next-stage expansion beyond Web and application servers to critical applications
  • Oracle Real Application Clusters (RAC), WebSphere, JBoss, real time, high availability, mainframe and massive scale
  • Be perceived as low risk, high value
• Simplify upgrade, maintenance and support, while adding functionality, expanded hardware compatibility and system management
• Drive toward a strategic data center position on virtualization

Scenarios:

• Most optimistic: Moves briskly to higher-tier Unix displacement
• Least optimistic: Noncommunity-spirited battles that recreate old Unix competitions of proprietary flavors
• Most likely: Overall, the Linux market eliminates barriers to entry; bifurcates into unpaid, internal or outsourced support or value-based support

Gartner has characterized the strengths, weaknesses, opportunities and threats (SWOT) on an aggregated basis across the three major environments: Linux, Windows and Solaris. It is possible for all three environments to take advantage of the emerging opportunities. Because of the compatibility offered on x86 platforms (chip-level feature/function enhancements), each of the vendors and environments brings a set of unique differentiators to the market.

Moreover, an OS is just one element of an increasingly complex stack of software elements. Interchangeability among vendor platforms is increasingly difficult — even though in one case, Sun Solaris, the OS is available on Sun x86 hardware, as well as on IBM, HP and Dell hardware. Applications used to be the major driving force in an OS decision. This will change with the increased interoperability of virtual machines across x86, and the abstraction of the OS and applications from the hardware. The impact will be that the value proposition moves up the stack, and the decision point will move with it.

The kernel richness will diminish in importance to the meta-OS features that enable the OS to function as part of a distributed resource environment. As IT resource planning becomes more complex in heterogeneous data centers, users will depend on the broader ecosystem to drive platform decisions. The return on investment gain will be crucial in the users' willingness to migrate off current platforms or expand their current platform usage. That decision's tipping point may hinge on just one or a few significant deliverables among the ones shown in Figure 7.