By Tom Groenfeldt

Coming soon to a company near you -- maybe to a competitor - services which aren't tightly coupled to data.

Actually, they have already arrived. Google is a prime example.

Rebecca Lawson, who is responsible for marketing and strategy for cloud initiatives and service management at HP, says this is an even more profound shift than cloud computing, but it hasn't really been recognized as a change.

Google is probably the best known service which runs against huge amounts of data that is not stored anywhere in particular, much less in a relational database.

"You can now store vast amounts of data and write services, not applications, against unstructured data," said Lawson.

The food industry could use this approach to provide traceability of products without the huge burden and expense of storing it all in databases. In the food supply chain, ingredients from multiple sources are brought together in a manufacturing plant. If all the manufacturers streamed their nonproprietary ingredient data to a single point, in the event of a food-related outbreak investigators could write a service against the data to locate the source of the problem.

"This will have a huge shift around how we think about data in the future. Google and Facebook are working on levels of data rather than on process." In the not so distant past, it was uneconomical to access large amounts of unstructured data, so firms built applications that were tightly linked to the data they used and developed processes and security around it.

Now data is growing exponentially because it is not constrained by an application infrastructure. The result could be many more of the unexpected correlations like the beer and diapers purchases uncovered by Teradata. Link to the case study.

Through data mining, it found a relationship between retail sales of beer and diapers on weeknights between 5 and 7 pm, presumably the result of young fathers stopping to buy diapers on their way home and picking up a six-pack for themselves as the same time.

A finding like that requires a massive relational database, which someone in a company has to fund and own, explains Lawson. While data mining won't go away, it is a very inefficient way to get at a single data point because it is serial. Cloud, by contrast, can run a huge amount of parallel processing against vast amounts of data.

"Cloud won't necessarily replace existing technology or business processing, but it will augment new values much like social networks are impacting businesses."

Comment on this article

Only registered users can write comments.
Please login or register.