Sanity Check: Downsizing, Rightsizing, and Fantasizing

by Bill Carico , ,
June 18, 2009

In 1993, when the mainframe was under heavy fire from alleged alternative platforms, I created a presentation titled “Downsizing, Rightsizing, and Fantasizing: Is Client/Server Really the Answer?” During the Total Cost of Computing (TCOC) section, I discussed my daughter’s first year at college. Though I had planned for tuition, books, and board, I quickly learned the total cost of college also included new clothes, a mini-refrigerator, dormitory accessories, bank account overdrafts, long-distance telephone calls, and extra trips home.

Also in 1993, U.K.-based consultant, Barry Graham and I conducted a seminar tour across 14 U.S. cities titled “The Downside of Downsizing.” This seminar featured a detailed cost comparison from Barry’s whitepaper “The Dinosaur Myth” that compared the cost per user over a five-year period to show the mainframe had the lowest TCOC compared to Windows and UNIX servers. Gradually, people began to realize that client/ server systems that supposedly cost less were actually 50 to 300 percent more expensive to own than mainframes when comparing the hardware, software, services, and most significantly, labor costs. This gap widened still further in favor of the mainframe if you added the costs for downtime and slower response times. Today, over a three- to five-year period, the mainframe typically costs about one-third of UNIX-based systems and a quarter of Windows servers when comparing the Total Cost of Ownership (TCO).

The mainframe’s economic advantages are realized when the system stays busy doing work. A modern mainframe is analogous to a jumbo jet: Although it’s the most expensive to buy, it’s also the most economical way to transport 400-plus passengers compared to smaller types of aircraft. This also is analogous to the school bus being more cost-effective than smaller vehicles. You see, it’s not about comparing Miles Per Gallon (MPG), per vehicle where smaller cars get five times more MPG than a school bus. It’s about the ratio of passenger MPG and passengers to driver(s). For a school bus, nine MPG multiplied by 36 passengers equals 324 passenger MPG per vehicle. An economy car that carries four passengers multiplied by 30 MPG equals 120 passenger MPG per vehicle. It would take nine vehicles with nine drivers to get the same payload (36 students) transported to school at the same time.

Comparing cost per user in computing is similar. For mainframe alternatives today, the people and maintenance costs can be 75 percent of TCO, while hardware and software are 25 percent. Thus, anyone choosing a platform based on acquisition costs is being myopic and will miss their budget forecast, or more likely, other projects will suffer because their staff has been repurposed to perform system maintenance. Compare the labor cost percentages on a mainframe and you’ll find that labor is 25 to 30 percent of TCO and the number of support staff required is shrinking each year. The mainframe has been automating workload management and systems management for more than three decades, and the result is less support staff can accomplish more compared to other platforms. I’m reminded of the UNIX/ Oracle debacle at a telecom company where they squandered almost $400 million trying to get off the mainframe. They eventually fired the IT execs overseeing the transition after only 10 percent of applications had been migrated to UNIX. At that point, they realized that twice as many UNIX system administrators than mainframe staff were needed to manage the 10 percent workload on UNIX vs. the other 90 percent.