z10: Data Center in a Box?
For many large mainframe customers, the idea of a mainframe-centric -- or mainframe-exclusive -- "enterprise data center" makes a lot of sense
It's difficult to overstate the significance of IBM Corp.'s System z10 mainframe, which constitutes a significant departure from Big Blue's Big Iron status quo.
To begin with, there are System z10's performance, power, and cooling, and virtualization benefits (see http://www.esj.com/enterprise/article.aspx?EditorialsID=3072). There's something more: another important refinement of the mainframe and its role in the enterprise.
For several years, IBM has been pushing the mainframe as an enterprise information hub. For its recent System z10 launch, Big Blue took the idea one step further, touting a vision of the mainframe as an all-in-one "enterprise data center."
What's so striking about Big Blue's pitch, experts say, is that it's got teeth: for many large customers -- and for almost all large mainframe customers -- the idea of a mainframe-centric (or mainframe-exclusive) "enterprise data center" makes a lot of sense. But what does IBM mean by this term?
It’s not exactly what you might think.
Big Blue doesn't envision a return to Big Iron homogeneity, experts say; instead, IBM is touting the mainframe as the best (that is, most ideal) platform for business resiliency, reliability, scalability, and adaptability.
That's what IBM means when it casts System z10 as an "enterprise data center" platform. "With the z9 family, [IBM's] umbrella message was about [mainframe] collaboration and integration. The z10 [Enterprise Class] hooks into some of that -- most notably the idea of providing business resiliency and security. However, the 'enterprise data center' that IBM is now discussing is more about [how an organization can deal most] efficiently with rapid growth going forward than it is with merely optimizing existing application workloads," observes Gordon Haff, a senior IT advisor with consultancy Illuminata.
This is a "change in emphasis," according to Haff -- not a fundamental shift on IBM's part. "It … reflect[s] a changing slant on some IT issues over the past few years," he argues, citing -- for example -- a shift in customer attitudes and expectations with respect to server consolidation.
"[W]e continue to see a great deal of interest in server consolidation, but, whereas the initial consolidation drive could have been summarized with the phrase 'Do more with fewer servers,' it's now part of a more complicated calculus involving rapid workload growth, simplifying operations, and running into power, cooling, and floor-space constraints.
In other words, it's not so much about cutting costs as it is keeping costs and complexity from growing exponentially."
This also jibes with what SHARE president Martin Timmerman says he's hearing. At the most recent SHARE conference -- which featured an actual System z10 running on the show floor -- there was considerable talk about the mainframe (and the new System z10 systems in particular) as a destination platform for x86 server consolidation efforts. But mainframe-centric server consolidation isn't a panacea, Timmerman says. Nor it is it something just anyone is considering.
The customers that embrace mainframe-based server consolidation are, for the most part, large shops that already have significant, multi-thousand MIPS investments. For these organizations, he stresses, the mainframe is an interesting -- and increasingly ideal -- platform for server consolidation not simply because of its processing power (which Big Blue touts as equivalent to about 1,500 x86 servers), but because of its array of exclusive features -- including integrated power and cooling capabilities; best-in-class virtualization features; a top-notch security feature set; demonstrable business availability, fault-tolerance, and business resiliency; and -- of course -- scalability.
In short, Timmerman argues, the mainframe makes sense as a server consolidation platform because it amounts to a data center-in-a-box. That's more or less the same pitch IBM itself is making.
"I certainly think [System z10] has a great consolidation story, and sometimes consolidations really work well and sometimes they're challenges. I think overall, we're seeing a number of consolidations, especially if people are going from Linux-based systems and Unix-based systems to run on zLinux, because the effort of any ports from those systems is not at all onerous," Timmerman argues.
"For large customers -- and SHARE has a number of large customers that attend regularly -- it offers more capacity, more CPU capability, and more memory capability," he continues. "Every data center is having power and cooling challenges, and to have this kind of computing power in one footprint with the increases in capacity without much increase in power requirements really means you can get a lot more CPU capacity [with] less power consumption."
Illuminata's Haff agrees. Thanks mostly to its best-in-class virtualization story, mainframe has become the go-to platform for hosting mixed workloads.
Any platform capable of simultaneously hosting a wide variety of workloads must also be highly configurable, Haff says, and, with its new System z10 release, IBM has significantly upped the ante for configurability.
The result, Haff maintains, is -- again -- a kind of data center unto itself. "[H]istorically, the mainframe was oriented around a relatively static stable of applications that would only go into production after a lot of careful up-front planning work," he comments, noting -- for example -- that changing the configuration of logical partitions (LPARs) used to require a system outage.
"As the mainframe evolved to handle a wider range of more dynamic workloads -- indeed that's no small part of its value proposition -- this sort of up-front crystal balling is less and less appealing," Haff continues. "With … [System z10], a variety of configuration changes can now be performed dynamically. These include adding or deleting LPARs, making changes to the Logical Channel Subsystem [LCSS], and changing the partitioning associated with logical processors or cryptographic coprocessors."
The value, Haff argues, is an all-in-one proposition -- an enterprise-data-center-in-a-box -- that's surprisingly (indeed, urgently) relevant today.
"The style of computing that the mainframe is most associated with -- [i.e.,] centralized and tightly-controlled -- is coming back in favor after a couple of decades during which more ad hoc decentralized computing styles proliferated wildly. At the same time, as the z10 EC makes clear, IBM System z also continues to leverage hardware and software advances and changes in both other IBM product lines and in the technology world as a whole," he concludes.
"The mainframe may have one foot far back in the past of computing, but it's also solidly grounded in the IT issues of today."