Mendelsohn sized up columnar databases (like Sybase IQ, Vertical and ParAccel) as suitable only for data marts. "Columnar databases have never been very successful beyond [marts] because you can't do updates efficiently; that's where they have had really poor performance," he said.
Oracle's answer is hybrid columnar compression, which gives the company "a lot" of the advantages of pure columnar, he said, including better compression and performance.
As for SAP's vision to use columnar technology for both BI and OLTP "some day," Mendelsohn dismissed the OLTP part as a research project. "If you look at what SAP is delivering with HANA, it's all around business intelligence, data warehousing and the traditional space for columnar storage," he said.
Reality Check. Sybase, Vertica and ParAccel have all been working hard on the slow-loading problem, and columnar databases have come a long, long way. These products are still not generally used for what you'd call an enterprise data warehouse, but they do deliver 100X or more compression versus Oracle's 10X.
What's more, the performance efficiencies of columnar querying are unmatched by any row-oriented database. It's a big, important niche in my book, though it should be noted that IBM, Teradata and Microsoft have also shied away from columnar. That makes me think that some sort of hybridization is in the offing.
As for Mendelsohn's take on what SAP has delivered thus far, he's right about it being BI/analytics focused, but he fails to mention the breakthrough of real-time access to transactional data. That's huge.
What about cloud-based databases like those from Amazon, Salesforce.com and Google, asked Rangan? (Another good question, though I would have mentioned Microsoft Azure as well.)
Once again, Mendelsohn's theme was "been there, done that." Oracle partners with Amazon and Salesforce, he said, and the Oracle DB and open-source Berkeley and MySQL DBs are all over Amazon's cloud and Salesforce.com.
While enterprise customers are intrigued by cloud databases, he said, they want to run their production systems on premises. This naturally led into great detail about what Oracle is doing to support private cloud deployments. Opening the door a crack, Mendelsohn allowed that everything developed for private-cloud development could also be quickly moved into the public cloud -- should circumstances call for that.
Reality Check. Why would Oracle, the biggest fish in the database market, want to hasten movement of databases into the (public) cloud? This doesn't support the mission of selling hardware (see below).
I imagine Oracle will be dragged directly into the cloud database market the same way it was dragged into database appliances -- years after competitors have already built a toe hold in a promising new market.
The Hardware Angle
What's Oracle's number-one opportunity? Here again, the answer was the Oracle Exadata Database Machine. In fact, whether the question was about data warehousing, OLTP, NoSQL, in-memory, columnar, cloud or whatever, everything seemed to get back to Exadata. That's because this product brings with it database licenses, related middleware and application software sales, and, perhaps most importantly, new hardware sales.
"We have 300,000 database customers and all of them have servers and storage running those databases," Mendelsohn said. "We're just scratching the surface of the opportunity to replace all those servers and storage. That's a huge growth opportunity."
Taking all these comments into account, the bottom line is this: Exadata is Oracle's hammer, and it seems to be looking at all markets and customer scenarios as nails.