Forrester analyst Jim Kobielus has predicted that data warehousing will evolve into a “virtualized, cloud-based, supremely scalable distributed platform.”

Greenplum, the massively parallel open source data warehouse company, says it’s already happening and that companies like Fox Interactive Media, Zions Bank and Future Group, the big Indian retailer, have already built early iterations of so-called Enterprise Data Clouds (EDC) using its latest widgetry.

It also figures that the Enterprise Data Cloud will displace the data warehouse appliance architectures that Oracle is so fond of, one of the reasons it’s supposedly buying Sun.

Greenplum claims that Oracle is already way behind and playing catch-up with its relatively new Exadata data warehouse appliance; Greenplum and Netezza have been offering appliances for years.

But it says hardware-based solutions such as Teradata and Netezza aren’t suited to the commodity hardware-based cloud infrastructure.

The cloud supposedly needs a software-only solution and that’s exactly what Greenplum’s got.

eBay, the world’s largest database, a hefty 6.5 petabytes, runs on Greenplum, which has collected 70 paying customers in the last two-and-a-half years. Netezza is supposed to have 200 customers and Teradata, the old man of data warehouses, has 900.

In Greenplum’s experience the 20-year-old mainframe approach of trying to create one single corporate-wide logical database is an idle and expensive exercise for a company to engage in.

Corporate units inevitably want things their way and so create silos – a psychological reality that the 10-year-old data warehouse appliance plays to – but then the data is fragmented and federated silos usually prove brittle when they aren’t having problems scaling.

The alternative is self-service, which means getting data into the cloud and out to the business teams as quickly as possible and letting analysts and DBAs instantly deploy all the data marts and data warehouses and run all the analyses on the data that they want.

Greenplum claims this “model less, iterate more” approach optimized for operations rather than performance and based on a common pool of physical, virtual or public cloud infrastructure (think VMware to start) is the right compromise.

Users get the control they want and IT gets to manage the pool as one infrastructure, increasing efficiencies and delivering predictable SLAs. Plus all the data, both the stable data and the volatile data that the mainframe approach invariably ignores, will actually be in one place.

Pieces of it will simply be broken off for any new warehouse without lots of process and upfront modeling; it’s supposed to be easy to share newly loaded data or analysis results.

The EDC approach implies elastic scale and massively parallel processing as well as a large-scale data collections and fast turnaround.

Greenplum’s new Database 3.3, now generally available, introduces key EDC features such as online warehouse expansion, which means it can be resized as needed across new servers added while the system is online and responding to queries.

Each additional server of course adds more storage capacity, query performance and loading performance.