In-Depth
Cloud Computing, Big Iron-Style
Mainframe pros have joined with IBM to tout the virtues of Big Iron-based cloud computing.
It's easy to dismiss cloud computing as just another case of vendor- and visionary-fomented hype far outstripping reality.
There are a few reasons to think that cloud computing's empirical reality (measured in terms of production deployments) will soon catch up with its undeniable hype (see: http://www.esj.com/Enterprise/article.aspx?EditorialsID=3252).
One big reason is just how seriously the big server OEMs -- Dell Computer Corp., Hewlett-Packard Co. (HP), IBM Corp., and Sun Microsystems Inc. -- appear to be taking cloud technology. Over the last few months, IBM has opened several cloud computing centers, adding to a quiver of computing-in-the-cloud resources.
HP, for its part, vaulted into the cloud fray earlier this year, touting a new Data Center Transformation services portfolio, with its cloud-based Adaptive Infrastructure as a Service (AiaaS) offering as the centerpiece. Through most of 2008, HP has been in the forefront of cloud computing: this summer it signed a deal with Intel Corp. and Yahoo Inc. to create a "test bed" service (the tautologically-titled HP, Intel, and Yahoo Cloud Computing Test Bed). Also this summer, Sun made its own belated entry into cloud technology, outlining a strategy to recast its utility computing division into a Cloud Compute service.
There's also scrappy x86 server stalwart Dell, which arguably beat HP, IBM, and Sun to the punch, unveiling its own cloud offering -- the Dell Cloud Computing Solution almost two years ago -- in March of 2007.
These and other efforts can be dismissed as so much public relations spin, of course. Less easy to dismiss, however, is a bona-fide push among seasoned technology professionals to harness the benefits of cloud in the context of the ever-available, ever-scalable mainframe. Mainframe systems, with their effortless ability to minutely manage resources, dynamically allocate new workloads (either by augmenting or bringing up new application resources or, just as likely, by firing up new virtual machine instances), and run a mix of proprietary (z/OS, VSE) and next-gen (Linux, Java) workloads, are seen as ideally suited for cloud computing. Now a growing number of mainframe professionals have joined IBM principals to tout the virtues of Big Iron-based cloud computing.
Take Jim Michael, a vice-president with the mainframe user group SHARE and an associate director at the California State University in Fresno.
Computing in the clouds isn't a far-fetched idea, Michael says. More to the point, he argues, mainframe-based computing in the clouds is a strategy that his own organization -- part of the California State University system -- is already pursuing.
"It is real already in some places, and it's obviously becoming capable in other places as time goes on," he explains. Michael, like other mainframe-focused cloud advocates, isn't so much a champion of external computing in the clouds -- i.e., a company purchases compute capacity from an external services provider -- as he is a proponent of an internal, Big Iron-based cloud model.
Big Iron-based cloud computing, which abstracts the hardware and software nuts-and-bolts side of the mainframe itself (abstracting, in effect, the perceived quirkiness that attaches itself to Big Iron) promises instant-on support for Java, Linux, or (in some cases even) classic mainframe workloads, Michael says. From an internal customer's perspective, it's just a question of buying capacity at a fixed price point and for agreed-upon service levels.
"You could have this model in which [the user says,] 'This is the OS I need. These are the requirements I have. Make me a server instance," he points out. "This is going to let us be much more nimble in meeting the user's requirements, because we don't have to go through the pyramid cycle to bring in the server or storage resources they need. We're architecting [instead] to bring these out of our aggregated resources," Michael continues.
There's a further bonus here, too, he argues: platform abstraction helps insulate against change, especially with respect to hardware underpinnings. "As I need to refresh this stuff, as I need to change it -- and I'm going to have to, it's just a fact of life -- in the cloud environment, because the user is disconnected from the actual hardware resources that they're running on, now when I refresh things on the backend, it's entirely possible that I'm going to be able to do that without the user knowing or caring that I've refreshed some of the components that they're running on."
Michael and Fresno State are currently experimenting with limited subscription-based licensing predicated (to a degree) on the very scheme he describes.
"Some of our cost models right now are subscription-based, where we can go to the end-user department and say, 'I'm not telling you now much it's going to cost you to buy the hardware, disk, storage, network -- no.' I can say, 'This is your ongoing cost. As long as you pay me this amount, you're going to get [the resources that you need] at this fixed price,'" he explains.
Michael says this is win-win for both business customers and IT. "You don't have to worry about the chunkiness of the curve. We can take a lot of the spike out of their [cost] curve and say, 'Here's your subscription model. You can pay on an ongoing basis, and we can allow you to do a lot better budget planning,'" he says.
Fresno State plans to move more ambitiously into subscription-based pricing over time, Michael indicates. "We're doing some of this already. I have a meeting [this week] with one of our end-user departments, and the cost model we're using there is a subscription model. We're doing the architecture for the systems that we're building and the architecture to support those systems, and these are the kinds of issues that we're talking about."
IBM itself is getting into the game. IBM officials have long talked up the mainframe as a locus for cloud, Web 2.0, and service-oriented architecture activities. Now Big Blue is putting its money where its mouth is, sponsoring a special session at the upcoming Winter SHARE conference (to be held in March in Austin, Texas) dedicated to dissecting the do's, don'ts, and why-the-heck-nots of mainframe-based cloud computing in the enterprise. According to Michael, the issue of cloud dovetails nicely with the much broader issue of virtualization, which will be one of two primary themes (the other being SOA) at Winter SHARE.
The mainframe-based cloud computing model, as envisioned by Michael and other proponents, begs an interesting question: inasmuch as it proposes to hide the mainframe (by exposing to customers common technology platforms -- e.g., Linux, Java -- and well-defined interfaces), is that, in fact, a good thing? Couldn't any platform be deployed in place of mainframe systems to accomplish a similar abstraction? Mainframe proponents such as Karl Freund, vice-president of global strategy and marketing, tend to downplay that suggestion, pointing to the mainframe's stellar resume relative to distributed systems.
"There just isn't another platform that's as secure, as scalable, as available, and as resilient as System z," Freund argues. "You take something like virtualization -- the mainframe was doing that long before any of these other [systems] came about. Virtualization on System z is best in class. No other [system] has the granularity, the scalability, [or] the ability to manage virtual [instances that] the mainframe has."
It's for this reason, Freund argues, that the mainframe is poised to play an important -- indeed, he argues, a foundational -- role in the “cloudification” of the enterprise. "Why wouldn't you make a platform like System z, which can securely and scalably host and manage these thousands of virtual [instances], the centerpiece [of a cloud-computing effort]?"
It's a trend that Freund expects to pick up steam. The cloud idea is still germinating, he says, and customers are trying to make sense of how best to harness it. Part of that process will involve separating the empirical reality of cloud from the metaphysical hype that (thanks to a series of high-profile, come-to-cloud adopters) currently shrouds it. Michael and other mainframe proponents believe there's empirical reality aplenty in the mainframe-based cloud vision. At the most recent SHARE conference held in August in San Jose, attendees got a cutting edge look at Big Iron in the Clouds.
SHARE vice-president Pam Taylor -- a solutions architect with a Fortune 500 company that she'd rather not name -- is one such believer. At San Jose SHARE, Taylor said, the idea was mostly to introduce mainframe professionals to the concept of cloud computing -- particularly as it relates (or might relate) to the mainframe. In this respect, she stressed, the focus was on cultivating an appreciation for internal cloud-based resources as opposed to the now-dominant external outsourcing model.
"Obviously, there are the early adopters, [who] are out there playing with it right now, but most of the world is trying to figure out what to do with it, what it means, what the value to them is going to be," said Taylor, in an August interview.
Since then, SHARE has expanded its focus on cloud computing; at the March SHARE show, IBM Research will lead a special session on cloud computing in the enterprise. Cloud computing isn't a cure-all, Michael concedes, but -- to his knowledge -- nobody's claiming that it is, either.
"No, it's not a panacea, but is it a player? Yes. Is it a big player? Yes. When you're looking at organizations that are offering cloud computing external to the enterprise, outsourcing to some extent, or rightsourcing some of your workloads into the cloud, there are people using mainframes to offer these services," he concludes.
"I'm more interested in the cloud inside the enterprise. One way I can deal with capacity requirements is to create centers of capability that function as a cloud to end-user departments that need to take advantage of the compute capacity but don't need to worry about acquiring the hardware or software. That's the most interesting enterprise use case, if you ask me."