IBM, Sun Prove Why Grid Computing is Alive and Well
Successful grid services from both IBM and Sun Microsystems offer proof that grid computing is alive and well in the enterprise
Are we on the cusp of yet another grid-computing hype cycle?
That’s not likely, and ,a pair of recent developments from IBM Corp. and Sun Microsystems Inc. show why grid computing is alive and well in enterprise IT environments.
Just last month, IBM introduced its Grid Medical Archive System (GMAS)—an industry-specific grid offering based on its Grid Access Manager. GMAS is one of several vertical grid offerings developed by IBM.
What’s significant about GMAS, analysts say, is that it extends an industry-specific grid program which IBM first outlined four years ago. In other words, industry-specific grid has earned its bona-fides: nay-sayers, who questioned whether grid could ever be applied to enterprise computing tasks, are being proved wrong.
More to the point, analysts note, grid computing has done so by ceaselessly innovating: IBM’s GMAS is based on a grid storage technology that few grid skeptics could have envisioned half a decade ago.
That’s the case for Sun Microsystems Inc.’s still-maturing Sun Grid service, too. When Sun Grid debuted two years ago, it was widely viewed as a not-ready-for-prime-time offering, in part because Sun was so secretive about just who was using the service and for what purposes. However, with 24 months of enhancement under its belt, Sun Grid now has the stamp of a mature—if less revolutionary—grid computing service, analysts say.
Add it all up and you have a few promising developments on the grid computing front—years after that technology rode the first wave of its technology hype cycle. "I believe that the rise in successful commercial offerings from IBM [and others] suggests that grid is definitely finding a place in the market. To be fair, I think that some of the early grid enthusiasts oversold the idea a bit. While it’s theoretically possible to create a grid out of nearly any IT infrastructure, it’s not something everyone would want or need to do," observes Charles King, a principal with hardware and storage consultancy Pund-IT.
IBM Continues Its Vertical Push
IBM positions GMAS as an integrated hardware, software, and services bundle for health care and life sciences (HCLS). GMAS’ strongest selling point is its virtualized storage capability, which makes it easier for health-care providers and life-sciences researchers to grapple with the storage resource management (SRM) problems posed by rapidly-expanding data volumes.
"Storage grids are the latest form of storage virtualization. They are beginning to take their place alongside, or even as an alternative to, other virtualization models that include host-based, … fabric-based, … and storage controller-based [technologies]," writes John Webster, a principal IT advisor with consultancy Illuminata.
Why do storage grids, in particular, appeal to HCLS shops? In part, Webster says, because they enable a kind of relatively painless storage virtualization. "That storage grids are decentralized by design differentiates them significantly from other alternatives. In fact, it is the decentralized nature of storage grids that allows storage administrators to think of managing their environments in very different ways—albeit for different use cases that focus more on archival storage, fixed content, and digital asset management—than as one-for-one replacements for high-scale primary storage," he indicates.
This kind of flexibility is tailor-made for HCLS shops, Webster points out. "[P]atient-care facilities are increasingly doing their medical imaging such as X-rays and MRIs directly to digital formats rather than analog film," he writes. "These files alone, when taken in aggregate, can consume hundreds of terabytes of storage, depending on the size and average age of the patient population served. Add to that the growth in the level of resolution and detail of these scans, as well as the growth of other types of patient files [such as] … documents, audio, video, invoices … [that are] also increasingly being committed to digital formats."
Storage Virtualization and Grids
Of course, HCLS isn’t the only vertical in which storage resources are taxed to the max. To the degree that Big Blue’s new GMAS service helps establish grid-based storage virtualization as a cost-effective alternative for HCLS shops, it might also pave the way for its adoption in other verticals, too.
"The GMAS solution specifically targets health care and life sciences, but its underlying grid storage architecture can readily be targeted at other vertical market segments. Finance and insurance, for example, also have significant digital asset and archiving requirements, as do entertainment and broadcast media, as well as various video surveillance, security, and satellite imagery applications," Webster concludes.
"Storage grids show how storage management practices must evolve to handle growth rates in excess of 50 percent per year coupled with mandates to preserve data in perpetuity. While IBM is not specifically positioning Grid Access Manager as general-form storage virtualization, it is, in fact, a grid-based implementation of virtualized storage. IBM is, in essence, extending its previous endorsement of grid architecture in computational environments to a production storage problem."
Pund-IT’s King salutes IBM’s vertical-specific grid strategy, which he says has enabled Big Blue to push grid computing for applications or scenarios in which it’s best suited—and not as a panacea for any and all enterprise ails.
"It’s … important to quantify and understand the financial and technical benefits before embarking on so complex an undertaking. The reason IBM succeeded in their vertical-specific approach is that industries—and their member companies—often utilize the same or similar applications and hardware, so developing replicable commercial offerings is less onerous and time-consuming than creating them one-off," King concludes.
Sun Grid Comes Into Its Own
After more than two years of gestation, Sun Grid—a pay-for-use grid computing service that Sun officials once compared to an electric power utility—seems to have come into its own, too. That’s in spite of a few fits and starts along the way, says Gordon Haff, a principal IT advisor with Illuminata.
"A variety of Sun executives cycled through Grid leadership roles and the state of Grid at Sun was long far less transparent than its price. Vague reference was made to large customers trying things out on a custom basis, but the much-promoted standard compute utility that anyone could sign up for with just a credit card was nowhere to be seen," Haff points out.
"As a result, other companies such as IBM and Amazon, ended up putting together more fully realized offers than Sun’s—even if they did so with far less hoopla." The bottom line, Haff indicates, was that Sun Grid never quite delivered on Sun Grid’s foundational premise—that customers could buy compute capacity (including both processing power and storage) for a single, transparent price.
Fast forward a couple of years, Haff says, and Sun Grid smells a bit sweeter. "What’s being offered is less ambitious than was once planned—both in terms of the types of services being sold and the markets they are intended to serve. But the fact that it’s no longer effectively vaporware has to be counted as a big step forward," Haff writes. "Today’s Sun Grid is only a compute utility. That’s always been the point of the arrow but storage was long part of the discussion as well. However, the ‘$1/GB/month’ concept isn’t part of the current offering, and isn’t on any public roadmap."
To be sure, Haff concedes, Sun does offer customers up to 10 GB of free storage per account (and customers can get additional storage on a case-by-case basis if needed), but he dismisses this as "scratch space" because the company doesn’t make any guarantees with respect to backup protection.
While Sun initially pitched Sun Grid as a jack-of-all-trades, it’s now focusing mostly on high-performance technical computing (HPTC) requirements—in other words, the kinds of applications in which grids have traditionally been dominant. What’s more, Sun’s initial vision of a vibrant Sun Grid ecosystem—replete with capacity-providing partners—hasn’t yet come to pass, either.
"Sun hosts its own Sun Grid infrastructure. Currently all the demand is being served out of a single datacenter in Nevada using about 1,000 CPUs," Haff notes. "Sun continues to maintain that it’s only doing its own hosting to prove the concept and work out the kinks; its preferred mode is to add capacity through partners as demand grows."
In the final analysis, Haff thinks Sun Grid might have been a victim of its parent company’s propensity for lofty visions and high-flown rhetoric. "Sun Grid has by no means yet achieved the revolution in computing that it initially promised. But it is now making progress. Sun sometimes seems to think that revolutions can happen overnight as the result of a blog post or a particularly dynamic launch event. They usually don’t, of course—especially when the revolution involves making deep-seated changes to the way that organizations purchase, deploy, and consume some of their most basic infrastructure," Haff concludes.