IBM i’s TCO Advantage Widens, According to Reports
September 13, 2017 Alex Woodie
IBM recently published a new report that claims the total cost of ownership (TCO) of an IBM i server over three years is dramatically lower than the equivalent Windows or Linux setups. It’s not the first such study IBM has done over the years, but what’s interesting is the gap appears to have widened considerably compared to a similar study six years ago.
In the latest TCO study, research and management consulting firm Quark + Lepton measured how much it would take hypothetical businesses to buy and manage three sets of servers: a Power Systems server running IBM i 7.3 and Db2; an X86 server running Windows and SQL Server; and another X86 server running Linux and the Oracle database.
With a three-year TCO of $430,815, the IBM i setup won hands down, according to the paper, which is titled “IBM i on Power Systems Meets the Diverse Needs of Midsize Businesses” and carries a publish date of August 30, 2017. (It’s interesting what one finds when one starts meandering through IBM Literature.) IBM i handily beat the Windows/SQL Server platform, which cost $1.18 million to buy and run over three years, and completely trounced the Linux/Oracle setup, which came in at $1.27 million.
The bulk of the differences in three-year costs among the platforms was directly attributable to personnel costs, according to Quark + Lepton’s report, which was paid for by IBM. Across the board, in each of the different hypothetical use cases the company drew up – including specialty retailer, discrete manufacturer, consumer products distributor, and agribusiness – the IBM i personnel costs were a fraction of the personnel costs for Windows/SQL Server and Linux/Oracle.
For instance, the personnel costs associated with running a modest Power Systems S814 server with an ERP/CRM application package for a consumer products distributor with 300 users was about $111,200, which assumed that only 0.3 full time equivalent (FTE) was needed. By comparison, the same software and hardware package on a Windows/SQL Server would require about $371,300, which reflected the higher staffing needs (0.5 FTE) as well as higher salary ($86,000 for IBM i admins compared to $100,600 for SQL Server database administrators). The Linux/Oracle setup, by comparison, cost over $450,000 in personnel costs.
The biggest setup in the report involved a hypothetical agribusiness with 5,000 employees and $1.65 billion in sales. The company needed ERP, procurement, EDI, and ecommerce capabilities to support 1,200 users. Quark + Lepton spec’ed out four eight-core Power S824 servers to run the various components, and one FTE. All told, the system would cost about $825,500 over three years. By comparison, the equivalent Windows setup would require 12 X86 servers and 2.5 FTEs, and cost $2.54 million, while the Linux/Oracle setup would similarly require 12 X86 servers, 2.0 FTEs, and cost $2.39 million.
“A number of fundamental core capability differences contribute to cost disparities,” Quark + Lepton, which is based in Boulder, Colorado, says in its report. “More granular partitioning and real-time workload management mean that greater workload density may be achieved with IBM i on Power Systems. Additionally, IBM assumes the responsibility for developing, testing, and preloading middleware components and applications, so customers don’t have to.”
We’ve seen these sorts of comparisons before. IBM previously worked with a group called International Technology Group (ITG) on a series of three-year TCO reports, the latest of which we wrote about back in 2011.
In that report, ITG concluded that the Linux/Oracle combination was 2.25 times as expensive as the IBM i setup, while the Windows/SQL Server platform was 1.73 times as expensive. For comparison’s sake, the Oracle/Linux setup was 3.3 times as expensive in Quark + Lepton’s August 2017 analysis, while the Windows/SQL Server setup was 2.74 times as expensive.
Quark + Lepton also analyzed the costs associated with downtime across the three platforms over three years and – not surprisingly– the IBM i platform won this comparison, too.
To figure out the cost of downtime, the firm calculated average cost of downtime for all six of its hypothetical companies “using appropriate industry – and organization-specific values,” it says. It then multiplied those figures by the estimated expected downtime for each platform, “based on user input.”
It projected that the three-year downtime costs should average about $935,500 for the IBM i server; $3.38 million for the Windows/SQL Server platform; and $4.46 million for the Linux/Oracle platform. According to Quark + Lepton, IBM i can be expected to have 72 percent lower downtime costs than the Windows/SQL Server platform and 79 percent lower downtime costs than Linux/Oracle.
One should take these TCO and downtime figures with a grain of salt. While anecdotal evidence has long suggested that IBM i offers a superior payback compared to other platforms – and, yes, better uptime to boot — there are many ways to encourage the figures to go one way or the other in these sorts of studies. The fact that IBM sponsored this would seem to indicate that the results, while not preordained, definitely were to Big Blue’s liking.
In the end, the data presented by Quark + Lepton are interesting but not earth-shattering. The report may be of interest to IT executives who are either pondering the elimination of the IBM i server or are considering it for a new business system. As for whether the numbers reflect the real-world, we’ll defer judgement. In the absence of independently validated testing and analysis of TCO of multiple platforms, this sort of analysis may be as good as we can reasonably expect to get.