The Future of Unix: Hazy and Overcast, So Reach for the Umbrella

Purely monolithic Unix will decline as infrastructure architects migrate legacy applications such as ERP off Unix; one scenario shows the period of sharpest Unix decline beginning 2014 to 2015.

• By 2017, 65% of applications running on proprietary versions of Unix in 2012 will have been migrated to x86 (primarily on Linux), making IT planning essential.
• ERP scale-out architectures on x86 platforms with Windows or Linux will rise from a 10% ratio in 2005 to 80% in 2017 for application and database management system (DBMS) online transaction processing (OLTP) workloads, making ERP/x86 standard.
• By 2020, the Unix market vendor share will comprise one predominant vendor, with about a 60% market share, and two others dividing the remaining share, incurring a potential procurement risk in a one-vendor market.
• Linux scalability and reliability will have surpassed Unix by 2017, but scale-out cluster and fabric architectures will remain the predominant Linux deployment style in 2017.
• The Unix decline will be steepest in the 2014 to 2015 period, when the heaviest migration from HP-Unix/Itanium is expected, and should be planned according to the circumstances at that time.

• For IT procurement decisions of a longer term (about 2017-plus as a time frame), allocate a declining proportion of the budget to Unix (no more than 20%).
• When planning a Unix migration exit strategy, make Linux your priority; or consider Windows if you're primarily a Windows shop with few Linux skills, and if strong expansion of Windows applications is anticipated.
• If you're planning one Unix vendor strategy, prepare proofs of concept (PoCs) and an alternative deployment on x86 as part of negotiations invoking either scale-out x86 (scale-up emerging) or scale-up reduced instruction set computer (RISC).
• Plan on hard partitioning, containers and high-density virtual machine (VM) deployment on scale-up x86 servers from 2015 to 2017 to satisfy demanding high availability (HA) high-end DBMS and application consolidation.
• Begin Unix migration plans in 2012 to 2014 for all applications except those with extreme uptime, latency and compliance requirements, and for those with sensitive client information/access (which can occur later, around 2017).
• Evaluate HP's Project Odyssey, and compare it with other vendor products for equivalent fabric and appliance alternatives in 2013 (i.e., HA/disaster recovery [DR], vertical scaling, uptime, partitioning, kernel health checks, pre-emptive failover, etc.).

We have mapped out four scenarios for the Unix market through 2017 – from least likely to most likely – with the factors that comprise the important driving forces in each scenario as our estimated probabilities, based on customer buying behaviors expressed to Gartner in confidential discussions.

Here are the four scenarios:

• Unix grows moderately (5% probability): Linux fails to provide the reliability and scalability needed for mission-critical, high-performance applications. Unix vendors deliver twice the performance in regular 18- to 24-month timelines. Intel fails to drive Moore's Law core performance. Red Hat encounters financial problems, and users lose faith in Linux. The anticipated HP-Unix decline is totally absorbed by an AIX or a Solaris migration. Oracle focuses primarily on SPARC Solaris.
• Unix growth remains flat (15% probability): Linux growth slows as users delay new workloads. Any x86 process/administration improvements fail to attract owners of business-critical workloads. Unix vendors deliver twice the performance in regular 18- to 24-month timelines. Intel fails to drive Moore's Law core performance. Red Hat is acquired by a software vendor for cloud purposes. Most HP-Unix migration remains loyal to Unix. Oracle manages a dual Solaris and Oracle Linux strategy.
• Unix growth declines moderately (30% probability): Linux growth continues to accelerate as ERP and DBMS users migrate to Linux. Independent software vendors (ISVs) polarize around a Linux or Windows development environment. Unix vendors consolidate to one or two key players. Intel continues to drive core performance. Red Hat and SUSE continue to accelerate their global market penetration. The x86 process/administration achieves operational parity with RISC/Itanium. Cloud infrastructure evolution favors x86. Linux wins the most Unix migration momentum. Oracle Linux gains momentum and market share in Oracle accounts.
• Unix growth declines steeply (50% probability): Linux growth continues to accelerate, as ERP and DBMS users migrate from Unix to Linux. Linux becomes the primary development environment (with Unix and Microsoft secondary). The Unix market collapses to one vendor. Intel continues to drive core performance. Linux-hosted virtualization accelerates. Cloud infrastructure investment almost totally favors x86. HP drives an aggressive "mission-critical x86" strategy embraced by other x86 vendors (including Oracle, as Exadata gains momentum).

Impacts and Top Recommendations for the Future of Unix

Source: Gartner (April 2012)

Gartner has seen a steady stream of clients migrating from Unix to Linux. Even when Unix retains a foothold after a major modernization or consolidation project, it is usually in a heterogeneous, rather than a homogeneous, workload role. While steady, the migrations have been phased by degree of complexity, total cost of ownership (TCO) savings, modernization priority and degree of internal Linux skills (or transfer of Unix to Linux skills). There is almost no reverse migration (Linux or Windows on x86 to RISC/Unix). Most of the applications on Unix are now consolidation environments comprising applications and databases, and exploit the high performance and latency of the RISC or Itanium backplanes.

However, several forces are acting to dislodge the Unix anchor:

• Xeon-EX (7500 series) technology with higher vertically integrated scalability
• Greater logical and physical scalability by the Linux kernel to within the range of Unix/RISC
• Maturation of hypervisors and virtualization tools for x86 to enable greater efficiency in the utilization of x86 servers, with broader adoption from users and support from ISVs
• Higher-bandwidth networking and clustering across multiple nodes

Moreover, the abundance of open-source applications that exist in the Linux world (e.g., the LAMP stack), the momentum of cloud computing on massively parallel infrastructures and the demonstration by vendors such as Oracle of ultra-high DBMS performance on racks running Linux are further attesting to the strength of Linux as the center of the technological dynamism occurring in the data center.

Recommendation:

• For IT procurement decisions that are of longer term (about the 2017-plus time frame), allocate a declining proportion of the budget to Unix (no more than 20%).

IT procurement departments are advised to carefully monitor inflection points that might result in maintaining a declining technology in a portfolio position. As a technology declines (see "IT Market Clock for Server Technology, 2011"), ISV enthusiasm will diminish, creating a shrinking ecosystem for the platform. Markets tend to create momentum effects in which fear and uncertainty drive decisions to continue to embrace once-popular technologies that are in decline. Organizations that linger in this mode often find themselves forced to "fence off" the legacy environment as one that's specific to a user group that cannot or does not wish to migrate, either due to the expense or to specially coded applications that cannot be duplicated on another technology without a high expense. It is important to limit such parts of the IT infrastructure to exceptions, that are not to be tolerated as a standard technology adoption and acceptance practice.

ERP applications and databases are mainstream to corporations, and require periodic refresh and enhancements to exploit the value they bring to the business. These applications comprise dozens to hundreds of modules and components. The two major ERP vendors, Oracle and SAP, and many business application developers for Windows have increasingly focused development efforts on Windows or Linux for x86. Their affirmation of x86 is a proof point that in their configuration testing they can readily meet the performance and scaling dimensions set by their applications' requirements.

By standardizing the hardware on x86 and compressing the number of OSs to two, ISVs help expedite system builds, upgrades and testing, and enable faster delivery of patches and enhancements with common management tools. It is these business-driven variables that create the driving force toward, and adoption of, the underlying platform. ISVs are also clued in on the technology road maps of the vendors. Without knowledge of the planned vendor investments and commitments to their technologies, ISVs would be hesitant to publicize their ERP directions. Occasionally, situations arise in which a strategic software vendor announces a road map change with an end of life to the development of software on a specific technology. Such occurrences would be less likely on standard, high-volume technologies.

Recommendation:

• When planning a Unix migration exit strategy, make Linux your priority – or consider Windows if yours is primarily a Windows shop with few Linux skills, and if strong expansion of Windows applications is anticipated.

This projection is a best estimate of where we believe market shares are headed. They arise from extrapolations based on current market conditions. Our quantitative analyses indicate that IBM is increasing its Unix market share, while the market share of HP and Oracle in shipped revenue is declining (see "Forecast Analysis: Servers, Worldwide, 2012-2016, 1Q12 Update").

We have no reason to expect a reversal of these trends, and continue to receive further affirmation of the trends from our client discussions. Unix will not be a target development environment for ERP, ISV and developers, although some in-house applications that are tied to the platform OSs and tools will be more difficult to transition to an alternative x86 environment.

Recommendation:

• If you're planning to have one Unix vendor strategy, prepare PoCs and alternative deployments on x86 as part of negotiations invoking either scale-out x86 (scale-up emerging) or scale-up RISC.

The advantages of RISC/Itanium in areas such as memory bandwidth and four- to eight-socket scalability will erode, because of the faster development pace of the x86 platform. For example, RISC processors introduced in 2010 will see little change before 2013, whereas x86 processors will see product generational upgrades every 18 months. Red Hat, SUSE and Oracle have all announced the latest improvements to the Linux kernel in production distributions now available. Their scalability has reached parity with Unix, leaving availability as the key metric to watch. With HP's intention to address availability to bring Linux up to the mission-critical standards of Unix in its Project Odyssey, even this frontier will fall.

Ironically, it was Oracle's decision to drop the future development of its software on Itanium that pressed HP to strengthen and differentiate its x86 servers. If successful, HP could create even more momentum around Windows and Linux for x86, increasing the erosion of the Unix market, on which Oracle with Solaris, as well as IBM, still depend. If Oracle's Exadata/Exalogic strategy (which is predominantly based on Linux/x86) continues to gain market momentum, Oracle's image as a champion of business-critical x86/Linux will rise. Similarly, an expected IBM April 2012 announcement can also strengthen IBM's perceived commitment to x86 as more balance to the Power and mainframe strategies.

Nevertheless, displacing RISC/Unix completely is no simple task, and we expect Unix to continue as a viable technology into the 2020 time frame. But these platforms will be in traditional enterprise data centers. The new hyperscale and massively hosted data centers that are already built or are in construction will be largely built as skinless, inexpensive, ultra-low-energy blades. Energy efficiency, footprints, disposability, massive scale and parallelism will outpace the monolithic styles of the previous decades.

Recommendation:

• Plan on hard partitioning, containers and high-density VM deployment on scale-up x86 servers from 2015 to 2017 to satisfy demanding HA high-end DBMS and application consolidation.

We again project and extrapolate migrations based on numerous Gartner client inquiries (in the many hundreds). Most users tell us that it makes little sense to "island hop" from one Unix system to another, or from one platform to another. In addition, it takes a great deal of effort to revamp scripts, tools and applications, and without a high degree of reassurance of success, they may again find themselves in a migration situation. By standardizing – and our clients believe that x86 and Linux (as open-source software [OSS]) are standards – users tell us that they have shifted to mainstream technologies whose stabilizing influence forms the large allegiance of independent hardware vendors (IHVs), ISVs and users like themselves. Such a gravitational pull will continue, as cloud and fabric infrastructures exert more influence in the latter half of the decade.

Gartner clients who use the Oracle stack and DBMS on HP-Unix/Itanium were initially unhappy about the Oracle decision to stop development of future generations of its software for this platform. However, they have slowly come to recognize the reality that they are on a timeline to eventual migration. HP tried to rally its user base by venting publicly about Oracle, insinuating that a lawsuit might reverse the Oracle decision. However, HP, too, has understood the migration reality, and made a decision to create a complementary strategy comprising flexibility and choice for users.

Since many HP-Unix users also run applications other than Oracle ERP applications and databases, the complementary x86 mission-critical strategy, side by side with the mission-critical Itanium/Unix strategy, with a potential convergence of both (in the same chassis), is what HP hopes will maintain alignment between users' interests and HP's technology road map for the longer term. Pure-play x86 vendors like Dell, Cisco, SGI, Hitachi and NEC will likely support – while seeking to gain their own market share – HP's efforts to stimulate market acceptance of greater x86 penetration (together with vendors like Fujitsu and Bull, with diminishing strategic Unix interest). Of course, Oracle's support for HP's x86 Linux and Windows ambitions on its x86 blade servers is an assumption. In any case, if HP is to win, then Project Odyssey must deliver on substantive benefits in cost, performance, availability and compatibility, while maintaining the value of Itanium/Unix if users prefer this technology (due to in-house skills, loyalty, migration costs and relationships).

Recommendations:

• Begin making Unix migration plans from 2012 to 2014 for all applications except those with extreme uptime, latency and compliance requirements, and with sensitive client information/access (which can occur later, around 2017).
• Evaluate HP's Project Odyssey, and compare it with other vendor products for equivalent fabric and appliance alternatives in 2013 (i.e., HA/DR, vertical scaling, uptime, partitioning, kernel health checks, pre-emptive failover, etc.).