For Orbitz, the cloud and Big Data fit hand in glove.

The online travel company recently began using Kognitio’s cloud-based data warehousing platform – Kognitio calls it Data-as-a-Service – to process and analyze large volumes of display advertising data collected from across Orbitz’s numerous web properties.

Specifically, Kognitio’s channel owners and marketing analysts tap Kognitio’s in-memory analytics database, called WX2, “to understand and stitch together customer segmentation profiling, the value of our display advertising and [its] impact on our customer base,” said Tony Gray, Director of BI Architecture and Operation at Orbitz. The company also links these insights with online site activity data and other demographics data “to paint a picture” of the user experience. The resulting analysis helps Orbitz determine its display ad placement strategy and where best to focus its marketing resources.

Kognitio offers WX2 — which supports business intelligence and online analytical processing on large, complex data sets – both on-premise and in the cloud. Orbitz chose the cloud-based option in part because of its speed of deployment, Gray told me. He said Orbitz had its Kognitio data warehouse up and running, fully loaded with its first batch of display advertising data, in less than two weeks. A similar initiative with an on-premise, traditional data warehouse would likely take months and require significant internal resources.

Since deployment, the data warehouse is updated daily via an encrypted FTP tunnel off Orbitz’s external-facing sites. As for security, “Kognitio is very rigorous in terms of encryption, data protocols and other safeguards,” Gray told me. Kognito’s security standards actually “surpassed our own internal requirements.”

Another important factor in Orbitz’s decision to go the Big Data-as-a-Service route was that by offloading the continued maintenance and related data integration and modeling tasks to Kognitio, Orbitz’s business users could focus exclusively on what they do best – analyzing the data to discover actionable insights.

“It really makes it a self-sustaining environment,” Gray said of the Big Data-as-a-Service model. “It’s a strong direction I think the industry is moving to.”

ServicesANGLE

I agree with Gray. From a technical perspective, the cloud’s elastic nature makes it an ideal place to deploy and manage Big Data projects, which often must scale-up (and occasionally down) as data volumes fluctuate. There is also a clear opening for service providers like Kognitio. As Gray alludes to, deploying and maintaining Big Data environments require highly skilled engineering teams and other internal resources. Cloud-based Big Data-as-a-Service providers like Kognitio remove these burdens and allow enterprises to start analyzing – rather than preparing – data in short order.

That said, Big Data-as-a-Service is not the be-all end-all of Orbitz’s analytics practice. The company also uses an on-premise EMC Greenplum database as a primary store for transactional data. Gray said Orbitz considered Greenplum for its display advertising analytics needs, but found Kognitio’s performance significantly better. He said users can query millions of rows of data and get results in under a minute with WX2, something not possible with Greenplum. (Gray and I didn’t have time during our conversation to get too deep into the technical details of why WX2 outperforms Greenplum, but I will follow-up with him and report back. In the meantime, I would love to hear from EMC Greenplum or other Greenplum users about this issue.)

Hadoop Compliments WX2

Orbitz also maintains a small but growing Hadoop cluster. It uses Hadoop to process site activity data that it receives in the form of unstructured log files. (Check out this presentation by Orbitz at last year’s Hadoop World for more details of its use of Hadoop.)

Gray stressed that he views Orbitz’s cloud-based Kognitio data warehouse and its on-premise Hadoop cluster as complimentary. Hadoop, he said, is a good, inexpensive platform for mining operation log files to better understand online site activity. But it takes significant expertise and technical resources to manage and to tease out those insights.

WX2, meanwhile, gives Orbitz’s business users a fast and reliable way to perform marketing-related analytics that they need to do their day-to-day jobs without needing to understand the underlying complexities of the system.

“Loading our log files into our Hadoop cluster is great, but the limitation there is its in a state where you need highly trained and highly skilled technical resources,” Gray told me. Together, Hadoop and Kognitio’s WX2 give Orbitz a more complete view of their business.

Time-to-Market Advantage

Going forward, Gray said he plans to investigate using Kognitio’s virtual cube tool called Pablo and hopes to add better data visualization capabilities (Orbitz currently uses both Cognos and QlikView on the front-end for data visualization.)

As for other enterprises looking to improve their analytic capabilities, Gray advises seriously considering a cloud-based strategy. Even if it’s technically feasible to deploy a large data warehouse or other Big Data technology in-house, the time it takes to do so could cost you. With a cloud-based solution, there’s the potential to be up-and-running significantly faster than if you built an on-premise platform.

Said Gray: “Folks need to really understand what their business users need and really understand the impact of time-to-market on these things.”