Mainframe computing received some interesting endorsements as summer drew to a close. At VMworld, VMware engineers took the stage to outline their vision of the future, citing mainframes as a kind of template or metaphor. But before you start thinking of warming up to those “V-partiers,” as I like to call the fan boys of hypervisor computing on x86 servers, let me give you my take.
I use the term V-party as a loose metaphor of the Tea Party that we have all come to know—and
love or hate—in U.S. politics. I don’t seek to get political here, but the parallels do seem to hold between the two. The V-party views itself as a “grassroots movement”—even VMware corporate evangelists say so on stage: “We are not a product, we are a movement!” However, as with the political entity, the V-party seems to draw considerable funding from corporate entities (EMC holds about 80 percent of VMware stock).
Both movements say they’re seeking to get back to basics and fundamentals, whether with respect to government tax policy in the case of the political group, or IT resource utilization efficiency in the case of the V-partiers.
Both say the status quo is a mess and that painful change is required to right the situation, and both claim to have the recipe—drawn in the case of the Tea Party from an interpretation of American Revolutionary history that doesn’t always fit the facts, and in the case of the V-Party from an interpretation of the history of the mainframe era that hardly fits the facts at all.
That’s the scary thing with regard to the V-Party embrace of mainframe concepts and principles. They’re showing a lot of nostalgia for a time that never really existed. They want to return to the homogeneous world of a mainframe-like computing stack in which one vendor calls all the shots—
provided that one vendor is VMware, of course. In their re-telling of the days of IBM hegemony in corporate computing, everything was shiny happy, all flowing with milk and honey. Only, of the folks saying this at VMworld, most weren’t even on this planet in the ’70s; they certainly aren’t old enough to have actually worked in a mainframe data center of that time.
They don’t remember how single-vendor hegemony was a two-edged sword, limiting in some cases the technology innovation, and in most cases, the cost containment of corporate IT. Mainframe technology was—and is—great stuff, to be sure. But it only came into full flower when the platform was opened up to third-party software developers and third-party networks and hardware. That’s when IBM’s innovation was reinforced and built upon by an industry, and when competition between IBM and IBM-compatible software houses began to bend the cost curve of mainframe computing.
Getting to this point in the evolution of the mainframe was a long and painful process that saw a lot of customer frustration. In the late ’70s, there was a grassroots, anti-mainframe movement that found its expression in “open systems computing”—a movement that also came from rebellion against the status quo, arguably funded by a handful of vendors, and ultimately one that delivered questionable benefits for consumers.
The V-party now casts itself as a break with the corrupt open systems model and a sensible return to more “mainframe-like” computing memes. They’ve arbitrarily added commands to the SCSI command language, their so-called “VAAI primitives,” without seeking the sanction of open standards broker, ANSI. They’ve added new storage APIs and now require all hardware vendors to embrace them if they want to connect to VMware-hosted applications. And soon, as announced in August, they will migrate all array controller functionality into their hypervisor software, adding more micro-kernels to that already bizarre collection of micro-kernels they term the operating system of fourth-generation computing.
The technical result of these changes is server-managed resources (just like those mainframes) that facilitate multi-tenancy and resource allocation dynamism (just like mainframes) with de facto standards that will bring unified and coherent operations back to IT (just like mainframes).
What’s missing? The stable core code base, for one—mainframes have one—and a sensibility about the role of open standards in facilitating competition and driving costs down, for another; IBM finally gets this.
Bottom line: What the mainframers pine for in their “mainframe” is the worst kind of conflation: one based on flawed history, flawed technology, flawed problem analysis, and self-serving solutioneering. That, I hope, isn’t at all “like mainframes.” I will pass on the V-party movement.