Last Thursday, EMC announced that it had spent $30 million to acquire intellectual property that IT services provider Acxiom had developed to provide middleware for grid technology implementations. Although grid computing has become something of a buzzword in the storage business and vendors such as Sun and IBM have offered grid technology for more than a year, customer demand for grid solutions has been weak so far.
It's not surprising that storage customers aren't at the forefront of users demanding grid technologies; storage vendors alone have been making lots of announcements regarding the direction of future storage-networking technologies, and grid computing, although it fits in this space, demands more from a customer than merely a commitment to a particular storage technology.
Grid technology is the key middleware for implementing utility-computing solutions, and in this space, you can't just make a piecemeal decision to implement the utility-computing model. This isn't to say that a business can't make a gradual, staged migration to utility computing, but it does mean that all parts of the organization need to buy in to the technology for it to work.
In a traditional storage-only upgrade, users (and the business process) don't really care where and how their data is being stored, only that the storage is fast, reliable, available, and secure. So migrating, for example, from server-based storage systems to NAS or SAN configurations doesn't usually require much interaction with end users or end-user applications because IT can hide the change from users by using fairly standard IT methods, such as drive mapping and virtual drives.
But with the utility computing model that EMC and other storage vendors are beginning to tout, the grid provides hosted services to the customer, adding capabilities, such as CPU cycles, on demand--and charging accordingly. This falls more in the realm of the ill-fated application service provider (ASP) model of the late 90s, when industry pundits claimed that provider-hosted applications were the wave of the future--a prediction that didn't pan out for the dozens of companies that disappeared trying to make the ASP space work.
However, I don't believe that the basic ASP concept was flawed. Rather, what killed the ASP marketplace was the lack of a reliable, secure infrastructure to deliver hosted applications on a broad scale to commercial end users. When that technical issue is overcome, the remaining barriers to ASP--and more recently, utility computing--tend to be more culturally oriented. This means that people need to be willing to cede their little islands of control to a more centralized infrastructure, which, of course, also has to prove its ability to deliver on demand. This is a much bigger hurdle than it appears since remodeling an ingrained management culture is incredibly difficult.
However, EMC is hedging its bets and has discussed making the grid technology available to customers in a nonhosted environment within the next few years. With this model, the grid technology would be completely self-hosted by the customer, with the vendor providing service and support. Although this grid-computing model might smooth the waters a bit for organizations considering the model for their future development plans, it conflicts with how vendors are currently marketing the technology. Namely, the primary advantages of grid computing are touted as, basically, paying only for what your business needs right now and being able to quickly deal with changing requirements. Self-hosting defeats this purpose, although conceivably an organization could build a large grid environment that contains sufficient virtualized resources to let it respond quickly to changes in demand without requiring the company to immediately buy additional hardware and software.
I'll be watching to see how grid technology places its own mark on the storage marketplace and what the big grid players do in response to these announcements from EMC.
