In representative installations in midsize manufacturing, distribution and retail companies, three-year IT costs for
use of IBM i 7.2 and Power Systems average 45 percent less than for use of Microsoft Windows Server 2012 and
SQL Server 2014, and 51 percent less than for x86 Linux servers and Oracle databases.
Figure 1 summarizes these results.
Figure 1: Overall Three-year Costs by Platform – Averages for All Installations
Costs included hardware acquisition and maintenance; license and support costs for operating systems, databases and other systems software; personnel costs for system and database administration; and facilities (primarily energy) costs. Hardware, maintenance and software license and support costs are based on discounted prices reported by users.
Installations are composites based on input from 42 midsize businesses employing IBM i on Power Systems, Windows servers or x86 Linux servers in manufacturing, distribution and retail companies. Companies ranged from $300 million to $1.65 billion in revenues, and employed 500 to more than 5,000 people.
Comparisons are between IBM Power S814 and S824 models equipped with POWER8 and dual- and four-socket x86 servers equipped with Intel E5 and E7 processors. IBM PowerVM, Microsoft Hyper-V and a comparable x86 Linux hypervisor are employed for virtualization.
Costs for use of IBM i 7.2 and Power Systems are lower across the board. For example, initial acquisition costs for hardware and software licenses average 35 percent less than for use of Windows and SQL Server, and 46 percent less than for x86 Linux servers with Oracle. Ongoing costs average 49 percent and 53 percent less respectively. Figure 2 summarizes these results.
Figure 2: Three-year Acquisition and Ongoing Costs by Platform – Averages for All Installations
A number of differences contribute to cost disparities. More granular partitioning and real-time workload management mean that greater workload density may be achieved with IBM i 7.2 and Power Systems.
In Windows and x86 Linux environments, separate servers are typically deployed to handle database, application and Web serving, and to support test and development systems. Hypervisors allow sharing of some physical platforms, primarily those supporting non-production and light-duty instances. The effects on overall configurations are, however, incremental.
In smaller installations, between three and five physical x86 servers are required for workloads handled by single Power System. In others, between 6 and 11 physical servers are required for workloads handled by pairs of Power Systems duplexed for redundancy. Hardware, software and support costs are multiplied accordingly. Greater administrative complexity also increases personnel costs.
The pricing picture has recently changed in favor of IBM i and Power Systems. New POWER8-based models offer better price/performance levels than their POWER7+-based predecessors. This is particularly the case for the company’s new four-core S814 model. Equally, Microsoft’s per core pricing model, first employed for SQL Server 2012, has pushed up Windows database server costs.
Moreover, per core costs for SQL Server and Oracle have been affected by the trends toward use of denser multicore processors. For example, where two-processor, eight-core servers might once have hosted core databases, the norm is now likely to be 12 or 18 cores, and high-end Xeon E7 processors embed 10 to 15 cores each.
Personnel costs are – by wide margins – lower for use of IBM i than for Windows and Linux servers. IBM i is the simplest, most automated, most tightly integrated operating system available today. It incorporates unique features such as an object-based architecture, integrated SQL database (DB2 for i), single-level storage and integrated workload management that minimize administrative overhead.
Similarly, because of the tight integration of DB2 for i, the same individual can typically handle database as well as system administration; i.e., separate database administrators (DBAs) are not be required.
Details of installations, along with methodology and assumptions employed, and cost breakdowns may be found in the Detailed Data section of this report.