DOWNLOAD FULL PRESENTATION HERE
Increasingly, businesses are trying to achieve better insights out of large amounts of streamed data by bringing it closer to real time processing power and querying from a live data mart. At the moment we are witnessing a convergence of workflows and technology platforms for real-time, analytics, cloud, and in-memory processing.
This allows us to effectively address time-sensitive business decisions that involve the volumes of big data while benefiting from the velocities of real-time processing. This involves real-time analytics capabilities for price optimizations, fraud detection, risk calculations, and operational business intelligence.
In this session, we presented a solution architecture combining volume and velocity dimensions of big data analytics utilizing both in-memory computing and a high-performance Spark distribution.
In the meetup, we examined a major airline who has implemented this solution for flight data yield management and optimization of seat inventory by forecasting against 300 million records and 2.2 million flight objects using Spark and an in-memory data grid. He also discussed how to reduce high latency associated with shuffling between HDFS and memory and reprocessing all data for what-if analysis.