Ingres: Vector Databases Are Real

In 2009, open source company Ingres was talking up its VectorWise database technology, promising orders of magnitude of improvement over current database products by harnessing processors better. This month, the technology has been launched, and the chief executive, Roger Burkhardt met up with eWEEK Europe to talk about databases and open source in general.

“Last year, Intel checked us out, and we demonstrated between ten times and eighty times performance increase in the labs, by unlocking the power of modern chips,” he said. “Those are kind of absurd numbers, right?”

Analytics just got cheaper

“What we didn’t know was could we perform the integration with real-world databases, and would it perform in the hands of customers?” he said, “because real world customers’ work loads are not benchmark workloads.”

“We are delighted,” he said. With over a hundred early adopters, the feedback is “absolutely extraordinary – the same figures we had in the lab.” The database solves problems that could not be solved before, and also offers simplification, he said.

The simplification angle was unexpected. “Performance is critical, but simplicity is also important,” he said, saying that cheaper analytics could produce a revolution simliar to when PCs brought spreadsheets within the reach of ordinary users.

One user had a 15 dimensional analysis, said Burkhardt, and ended up using a single table 350 columns wide: “He was amazed how fast it worked.” And it’s also faster to develop and deliver this way, he said.

At present, users have to “wait till the high priesthood have built special schemas” before they get analytics queries run. With VectorWise, they should be able to run those queries at will, he said. “This really is an industry breakthrough – we see nothing else like this in the market.”

By making analytics cheap and quick, it also enables a “new class of analytics applications”, where the analytics is embedded in the application, instead of sitting in a paper report on a desk somewhere.

It also allows companies to put off big purchases of database hardware – it promises to compete with the expensive stuff but using general purpose hardware. “We’re focused on the mid-tier of the market – datamarts holding one to ten terabyte of data.”

It’s also going to have a green message, since it reduces the amount of hardware (and in paritcular disk drives) but he doesn’t have benchmarks yet. One case study has Ingres VectorWise replace an eight processor, 144 disk system, with a two processor system using four disks. “There’s a big opportunity to reduce costs,” he said. “There is an awful lot of ‘busy work’ going on in the servers.”

Where’s the competition?

If Ingres is really ahead in this, how long until rival companies catch up, we asked?

“This requires a fundamental  rewrite to the core of your product,” said Burkhardt. “The research goes back to 2004. I actually hope the rest of the industry catches up over time, because this is a good step forward – but we are not standing still.”

“I think it will come step by step over a number of years,” he said, “and the most immediate response will be marketing announcements where people pretend that they have it, by applying some slick gloss to some rather thin work.”

He predicts a lag, “while they decide whether the competition from people like us is enough to force them to go do this,” but also suggests companies which sell hardware might not be so keen to bring the technology to users.