Intelligent Enterprise

As I wrote last week, the information available on the HP Oracle Database Machine and HP Oracle Exadata Storage Server is incomplete. Pertinent questions are on the table, but I've been unsuccessful, thus far, in getting answers. Maybe the right people are taking a few well-deserved days off after Oracle Open World or maybe they're observing Rosh Hashanah. Nonetheless, I've not heard back on requests made Friday and again yesterday. I have, however, talked to HP about the fit between this new product and Neoview.

To review, the key questions about the HP-Oracle offering get into the nuts and bolts of the hardware. Is it shared-nothing architecture through and through, and, if not, how does the optimizer negotiate between the two sides of the devise (database and storage)? Second, if it's built on "industry standard" hardware, just how does it put query processing power "on each and every disk," as suggested by Larry Ellison? Netezza, for example, gains its performance by putting query processing power on a Field Programmable Gate Array on each and every disk, but that's proprietary hardware.

I'm still waiting to hear from Oracle on these questions. In the interim, IBM declined to comment on Oracle's news and HP referred all questions about the joint HP-Oracle products to Oracle.

So what about the fit with Neoview? Ben Barnes, vice president and general manager, Business Intelligence, HP Software, made it clear that he and his group are were not involved in any way with the development or positioning of the new HP-Oracle offerings. Here are a few of his comments:

• "My understanding is that one of the main functions and features we're seeing on the new HP-Oracle appliance is its ability to load data without creating indexes, material views of other types of structural partitions and then be able to scan that data at an incredibly fast rate. That capability would lend itself to data warehouses and large data marts, but in my opinion it's not designed to be accessed by thousands and thousands of concurrent users because you're going at [the data] in a full table scan type of approach.

• "We believe there's a position in the market for Oracle's machine, and there's a position in the market for Neoview. Neoview is designed for enterprisewide deployments where you have not only large data sets, you also recognize the need to create indexes and structures to manage the data. You'd want the capability to manage thousands if not tens of thousands of concurrent users. You'd want to be able to service both quick response times and at the same time be able to run, in a parallel fashion, large, complex queries. One of the clear differences [with the HP-Oracle Database Machine] is that it's still the Oracle 11g RAC database, whereas our database is based on a shared-nothing, massively parallel architecture. There are some basic, fundamental differences between those two architectures."

That brings me full circle back to the architecture questions. Going back to the Oracle Open World presentations and available press materials, the thing that strikes me is that all the comparisons seemed to be against conventional Oracle or conventional competitive database deployments. As impressive as 10x, 20x or 50x performance may sound, I didn't see any comparisons, at least as described by customers, to Teradata, Netezza, Neoview or other shared-nothing, massively parallel processing deployments and performance benchmarks. It appears we'll have to wait and see…