What's behind Teradata's Hadoop volte-face?

Data warehouse giant Teradata has ridden a wave of hype with Hadoop. But now it says customers should jettison the poster-child of the big data boom. Lindsay Clark asks why.

Six months ago, in Madrid, Teradata Chief Technology Officer Stephen Brobst said companies considering migrating away from Hadoop were "throwing the baby out with the bathwater". 

"They actually shouldn't want to get rid of Hadoop… they should rationalise it, so they're using it for the right things, not the wrong things," he said.

But at Teradata's Universe conference in Denver last month, none of the data warehousing company's executives had a good word to say about the distributed file system which became a cornerstone in the multi-billion-dollar big data boom. It was cumbersome and costly to maintain, they insisted. Customers were walking away from hundreds of millions of dollars of investment in Hadoop systems after experiencing poor returns, Teradata CEO Oliver Ratzesberger said. A plank of Teradata's new strategy is to offer migration away from Hadoop.

So why the sudden change of heart from Teradata? Technology moves on, but it is not changed radically in the last six months.

Many of Teradata customers at the conference were happy to continue with Hadoop playing an important role in their data architecture for the foreseeable future.

Hadoop works for us

For example, NXP Semiconductors, a $10 billion microprocessor manufacturer, has completed a three-year migration to Hadoop as the main ingestion for its data analytics and machine learning environment. Data scientist James Bird told the conference: "We've been moving a lot of our data streams into an on-premise Hadoop environment. The Hadoop is a complex ecosystem, and there are some challenges we have run into that we did not necessarily anticipate, but it is fulfilling the needs that we have today. Ten years from now, will we still be using Hadoop on-prem? Probably not: we will have probably moved over into something like AWS S3. But today the Hadoop environment works for us."

To continue reading this article register now