In-Depth
IBM Announces Ambitious Distributed-to-Mainframe Consolidation Project
At $250 million over five years, the projected savings are real and substantial -- and the PR benefits, especially for System z, are even more important.
When it comes to cutting data center power and cooling costs, IBM Corp. practices what it preaches.
Big Blue last week announced an ambitious server-to-mainframe consolidation project, positioning it as part of its Project Big Green push. (http://esj.com/Case_Study/article.aspx?EditorialsID=2615) Over the next two to three years, officials say, IBM plans to consolidate nearly 4,000 distributed servers onto just 30 System z Enterprise Class mainframe systems.
The projected savings, says David Gelardi, vice-president of IBM industry solutions, are real and substantial. Gelardi and other IBM officials point to sizeable energy savings—the new mainframe farm is projected to consume 80 percent less energy—and huge attendant cost savings. "It’s hard to get a good handle on [the savings]," he concedes, "but I will tell you it’s … certainly well in excess of three-digit millions of dollars. We’re still trying to figure out exactly what it’s going to be, in part because we’re planning [the consolidation]."
Officially, IBM anticipates saving as much as $250 million over five years. The bulk of these savings—real or imagined—should stem from the elimination of 3,900 individual servers, each with a power cord and a power supply, says Gelardi. "The savings in space, the savings in power, are fairly obvious. It’s 30 refrigerator-sized things versus 3,900 other things," he points out. "[Additional] savings will come from a couple of other capabilities. The mainframe itself has a long history of [being able to] manage multiple workloads. Virtualization technology has been built into the mainframe practically since day one. And Linux itself has matured. As an OS, Linux has achieved practical ubiquity."
The servers that IBM plans to eliminate constitute about 45 percent of its worldwide inventory of 8,500 servers. "We’ve targeted a little less than half as appropriate for consolidation on these new mainframes that will be rolled in, and the 3,900 servers will be moved out, or appropriately resold, refurbished, or harvested," he indicates.
What makes a server an "appropriate" candidate for consolidation? For starters, Gelardi says, it’s typically running a Unix or a Linux workload. On the other hand, he says, there are cases in which Win32- or Win64-based systems can be candidates for consolidation, too.
"If you look at our internal usage of software, we have over time moved fairly aggressively to a DB2/WebSphere interface, so that helps us in the migration. You have this layer of abstraction between the application code and the system, [so] it almost doesn’t matter what the underlying microprocessor and machine are," he says. "When we looked at the 3,900 consolidation target, the vast majority of them are either Unix or Linux [servers], with some Microsoft [systems] where we have great portability or [where the] software provider has the ability to support a Linux environment."
Nor does IBM eventually plan to consolidate all of its distributed server assets onto mainframe hardware. "Although we’re doing a massive consolidation and we’re taking out 3,900 [servers], I do not want the impression to be left that every single application in the world is a target for this form of consolidation," he warns. "We continue to deliver our POWER-based solutions, both AIX and i5 OS. We continue to have Intel and AMD servers. We have specialty products like BlueGene. Really, every workload has a different set of requirements."
Industry veteran Charles King, a principal with consultancy Pund-IT, sees Big Blue’s move as integral to an overall greening trend across the industry.
"A mere couple of years ago, discussions of energy-efficient technology were largely limited to low-power microprocessor advances and specialty products such as add-on liquid-cooling solutions for server enclosures," he points out. "Today, energy efficiency is a key point in solutions, ranging from increasingly mundane server virtualization and consolidation strategies to leveraging 3D modeling technologies for data-center design." King says the big problem is one of design and implementation. In this respect, he sees IBM’s effort as a proof-of-concept for other would-be greening initiatives.
"Replacing 3900 servers with 30, even if they are mainframes, will significantly reduce overall power consumption, server footprint, and data-center cooling requirements, resulting in lower electricity and facilities costs. In addition, IBM should also save a considerable sum on software licenses and better manage its IT administrative staff," he points out. "The new mainframe infrastructure will also provide the company a means to easily increase performance by adding virtual Linux servers as needed, an interesting demonstration of the profound connection between IT flexibility and overall performance." On the other hand, King describes IBM’s projected cost savings of $250 million over five years as "small potatoes to a company of IBM’s size."
He believes the intangible benefits of Big Blue’s move might actually trump cost savings, simplified manageability, or other more tangible benefits.
"First, the effort offers IBM a public means to highlight capabilities of its homegrown technologies, particularly its signature mainframe platform," he concludes. "Related to this, the consolidated data centers will provide the company living laboratories for demonstrating the process and benefits of IT infrastructure consolidation to interested customers. Finally, and perhaps most importantly, the effort highlights IBM’s innovative approach to data-center energy efficiency and the obtainable end-to-end benefits of Project Big Green."
See Related Articles
About the Author
Stephen Swoyer is a Nashville, TN-based freelance journalist who writes about technology.