Resources
Resources
Explore our expanding library of Big Data resources.

Airpush Case Study

Case Study

Airpush Accelerates, Expands Analytical Insight with Altiscale

Named “Top Innovator in Advertising Data” at the 2014 DataWeek Conference and ranked #2 in the 2014 Forbes list of “Most Promising Companies,” Airpush is on a mission to redefine mobile advertising. Its self-serve advertising platform, *AirDSP**™*, empowers clients to purchase inventory from the Airpush mobile ad network and all major mobile real-time bidding (RTB) exchanges. Airpush’s exceptional ad formats and Hadoop-driven targeting technology are delivering the industry’s best performance to 150,000 mobile applications and the world’s leading advertisers.

Challenge

Airpush’s innovative mobile advertising targeting technology continually analyzes vast amounts of data in order to determine how advertising content, format, and placement impact performance. With the resulting insight, Airpush optimizes advertising placement and performance, as measured by metrics such as click-throughs and conversions. Initially, Airpush housed advertising data in four different, siloed applications, slowing analysis and time to value. To streamline and speed its analysis, Airpush made the decision to move the entire process to Hadoop—which provides a single, low cost, large volume data store. After deciding to transition to Hadoop, Airpush began its search for an effective platform on which to run it.

A Comprehensive Platform Purpose-Built to Run Hadoop

In its search for a Hadoop platform, Airpush tried Amazon Elastic Map Reduce (EMR) and quickly realized that its reliability, scalability, and support were insufficient. After evaluating a number of additional options, Airpush selected Altiscale’s big data platform, based on Hadoop and Spark, for its high performance and scalability, outstanding technical and end user support, low total cost of ownership, and ability to run increasingly complex analytical models. The Altiscale Data Cloud, with petabyte-scale infrastructure purpose-built to run Hadoop, significantly reduced job failures, leading to fewer retries and faster job completion.

Unlike cloud providers that offer only minimal support, Altiscale monitors and manages its service around-the-clock, proactively addressing issues that customers run into in production—such as resource contention, performance optimization, and infrastructure upgrades. This superior support, combined with the high performance of Altiscale’s innovative, industry-leading technology, dramatically decreased expenses Airpush incurred due to the maintenance and management of Hadoop. “With Altiscale, we don’t have to devote a team of engineers to managing our Hadoop. This not only frees our internal resources to focus on core business issues, but also substantially lowers our overall, Hadoop related costs,” said Aditya Chandra, Vice President of Infrastructure, Airpush.

Benefits: Lower Total Cost of Ownership, Increased Expertise and Support

Now that Airpush has transitioned its advertising data and analytics from siloed applications to Hadoop, it’s gaining insight in a fraction of the time previously required. By implementing Altiscale’s comprehensive big data offering, Airpush has also lowered its total costs while increasing its Hadoop reliability and scalability—and gaining ready access to extensive expertise and support. “When we were using Amazon EMR, our questions were met with standardized, emailed responses that simply didn’t provide the level of support we need,” said Chandra. “Altiscale responds to our inquiries with real answers given by real experts, to address each specific situation. Their support has consistently risen above and beyond our expectations.”

Airpush plans to further leverage Altiscale’s robust capabilities by moving additional analytical processes to Hadoop and utilizing Spark to increase the complexity of its analytical models. Due to Spark’s full integration with the Altiscale Data Cloud, Airpush has been able to accelerate the timeline for this transition. “With Spark running so smoothly as a part of Altiscale’s platform, we’ve been able to move forward faster than planned,” added Chandra. “As we extend our analytical models and transfer more of our processes to Altiscale, we expect to further decrease overall costs and speed time-to-insight. Altiscale’s big data platform is providing us with a clear path forward, toward a future of expanded analytical insight and continued business success.”