Platfora has introduced what it describes as the first in-memory business intelligence platform for Hadoop. The company unveiled its product, which is now in beta, at the Strata + Hadoop World Conference in New York. It will go GA in Q1 2013. “There is a lot of excitement out there among our early customers as well as the Hadoop distribution companies and other systems vendors,” Ben Werther, founder and CEO, Platfora, tells 5 Minute Briefing. "They all recognize that Hadoop is this incredibly promising technology," he notes, but having to build data marts in between and have connectors is a bottleneck, and so they are "very supportive and see us as complementary."
Platfora enables users to change their business intelligence questions without involving teams of IT staff to re-organize a data warehouse. Because it allows users to determine the questions after data is collected, the company says it essentially eliminates the need for a data warehouse or ETL (extract, transform, and load) software.
Werther likens Hadoop to a “huge cardboard box” that you can throw all the data at, but then using that data is difficult. The conventional wisdom today says you need to hire data scientists who can sift through the box of data, but, he says, “The problem isn’t finding those developers; the problem is that the platform technologies are missing. And so what we do is we complete the stack above Hadoop in a way that everyday business users - product managers, finance people, marketers - can interact with that data.”
Platfora transforms raw data in Hadoop into interactive, in-memory BI without the need for a traditional data warehouse, and end users are presented with a web-based analysis interface using HTML5 technology. Platfora works by pulling data out of Hadoop into an in-memory data processing engine.
The software installs on one or more servers that sit next to a Hadoop cluster. It interacts with the Hadoop cluster via standard interfaces and Platfora supports all of the major distributions, including Cloudera, MapR, Hortonworks, and Amazon’s Elastic MapReduce. “Whatever Hadoop distribution the customer is using, we can point at it directly,” says Werther.
According to the company, Platfora’s biggest innovation lies in how the layers work together seamlessly. It automatically refines the in-memory data based on the questions being asked by end users.
“For dimensional data, hot plugging out to customer systems of record and things like that, we support doing that automatically via standard ODBC and JDBC-type connections and that is all relatively small data on the side,” says Werther. “For any of the big data aspects of this, the way this is trending is that all of the data that going to those other systems is increasingly also being fed into Hadoop - that doesn’t mean that those other systems will go away, but it means that the customers we are talking to understand that they can push that data into Hadoop as well and they can be deriving value from that data using our tool and our platform and not have to do data integration and ETL and all the custom processes that they have to today. We think that is a much cleaner approach.”
In fact, he says, “Most of the data we see is a mix of truly structured or semi-structured data like log records or JSON records and so on. There is obviously unstructured data that goes into Hadoop but I think people underestimate the ability to put real structured and semi-structured data in there and drive interactive visualization and BI from that in a way that works as well and is much more flexible and more rapid to deploy than trying to go model it out and do it in the traditional relational way.”
According to Werther, unless businesses can explore and interact with their big data in a timely manner, the promise of Hadoop is lost. “What we do that is different from anyone else out there is that we transform Hadoop from this slow batch engine into a platform for sub-second interactive business intelligence and analytics,” he says. “Literally within hours of installing our software, these business users are building dashboards, building visualizations, exploring and deriving insights from all of this data that they have collected in Hadoop in ways that were not possible before. And there is nothing out there that does this today in anything close to the way we do."
Platfora is funded by Andreessen Horowitz, Sutter Hill Ventures and In-Q-Tel. For more details, visit www.platfora.com.