IBM’s Energy Efficiency Smoking Gun

Industry watchers see Big Blue’s mainframe power efficiency push as a signal event in the longstanding scale-up versus scale-out system wars

IBM Corp. raised more than a few eyebrows when it announced a new mainframe power saving program—a so-called "mainframe gas gauge"—for its System z platform. Analysts found merit in Big Blue’s mainframe power efficiency concept (see

In fact, more industry watchers say, IBM seems to have thrown down a gauntlet of sorts to the rest of the industry: put up or shut up.

To recap, Big Blue announced a new watts-per-unit energy-consumption metric that it says demonstrates a consistently high utilization rate for its System z mainframes (see

IBM also promised to start publishing typical energy-consumption data—derived from in-the-field monitoring of approximately 1,000 customer machines—for System z. IBM officials said the company’s latest Project Big Green gives customers an easy-to-understand energy rating that can help them determine just how energy-efficient System z actually is.

Sound familiar? IBM’s David Anderson compared his company’s watts-per-unit metric to another prominent industry standard, the ubiquitous "kilowatts-per-year" metric used to rate the efficiency of refrigerators and other appliances. Anderson argued that Big Blue is measuring average watts/hour consumed, which can in turn be used to calculate watts per unit. The result: a kind of mainframe gas gauge.

"The customers come from all industries and geographic areas. Being on z9, they are exploiters of the latest mainframe technology. We have the ability to go min, max, as well as typical. Min and max are already available using the power estimator. Few configurations resemble those," he said.

Several industry watchers have praised its mainframe-centric Project Big Green push—among them Gordon Haff, a senior IT advisor with Illuminata. In a research bulletin published two months ago, Haff cited IBM’s recent distributed-server-to-mainframe consolidation effort as one reason why Big Blue’s Project Big Green push (and, especially, its mainframe aspect) deserves to be taken seriously.

According to IBM’s own projections, it will migrate almost 4,000 distributed servers to 30 high-density (i.e., 54-way) z9 mainframes at six data centers scattered across the globe. Haff says Big Blue’s migration effort should result in its shifting a substantial proportion of its application workloads on to the mainframe, and that is telling.

"An interesting aspect of this server consolidation is that today only about a third of IBM’s applications are running on System z, and those only represent one percent of their total physical servers," he pointed out. "From the remaining two-thirds, much of the software runs on platforms such as DB2, Domino, WebSphere, or Web servers."

It’s probably coming it a bit too high to suggest that IBM can simply shift these workloads over to System z or z/OS with the flip of a switch, Haff conceded, but neither is such a migration an altogether daunting prospect. "While it would be a bit flippant to suggest that these sorts of applications can ‘just move’ from one platform to another—minor tweaks will often be needed to handle version differences or even just changes in physical location—neither will they tend to have particularly deep ties to a given operating system or server architecture," he points out, adding that IBM looked at about 16,000 server images during the planning phase of its consolidation project: "They determined that it didn’t make sense to move about half; the other half were ‘possible.’"

Other analysts are positive about IBM’s mainframe-centric evolution of Project Big Green. Industry veteran Joe Clabby, president of Clabby Analytics, is the latest and, perhaps, most enthusiastic. "IBM now has ‘the smoking gun’ when it comes to proving how scale-up architecture is superior to scale-out architecture when it comes to energy efficiency," he argues. "With this data in hand, enterprises can now perform comparative analyses that will make it possible to understand the true benefit of moving away from distributed, rack and tower computing designs to more energy efficient scale-up designs."

Clabby sees the availability of Big Blue’s mainframe "gas gauge" as a signal event in the longstanding scale-up versus scale-out wars.

"If CEOs, CIOs, CFOs, and information technology executives take the time to understand what this data really means—and if they extrapolate how much energy could be saved in their respective computing environments—they would all be hard-pressed to justify the use of scale-out computing designs to their stockholders, much less to the rest of us who want enterprises to stop wasting greenhouse-gas-producing energy."

Just how efficient are IBM mainframes in this regard? According to Clabby, a single zLinux implementation can perform as much work as 250 x86 processors—while at the same time consuming as little as two percent (and as much as 10 percent) the amount of energy.

That’s not all. Clabby notes that Big Blue’s data demonstrates that typical mainframe energy use is normally about 60 percent of a model’s "label" (or maximum) rating. In other words, he says, mainframes are actually performing about 40 percent better than previously rated.

"In an age when energy prices are going through the roof, and where there are real concerns about environmental damage caused by the fossil fuel many states use to produce electrical power, this kind of message should resonate throughout the datacenter," he points out.

Clabby, like other industry watchers, bemoans the "schlocky" aspects (e.g., mainframe "gas gauge") of Big Blue’s marketing effort. If you look past the schlock, however—and if (for the Big Iron-phobic) you can also swallow the "mainframe" part of IBM’s announcement—Big Blue has actually precipitated a compelling (and customer-friendly) shift in the enterprise sever segment.

"What IBM has implemented is a monitoring/measurement system that clearly shows how much energy can be saved by using scale-up systems designs," he concludes. "IBM should take this program a step further — and compare and contrast its four distinct server lines across various workloads. Standards should also be developed using this technique to compare the servers of Sun, Dell, and Hewlett-Packard with those of IBM."

About the Author

Stephen Swoyer is a Nashville, TN-based freelance journalist who writes about technology.