VISUALIZATION

Cluster Computing Helps Farecast.com to Manage and Mine Terabytes of Data Critical to Providing Accurate, Timely Airfare Information to Consumers: This holiday season, hundreds of thousands of business and vacation airline travelers will turn to Farecast.com, the airfare prediction website, to determine the best value and lowest airfare for their trip. Launched in June 2006, the company's patented airfare prediction service shows whether the lowest fares for a customer's trip are rising or dropping. Currently, Farecast.com features free airfare predictions for travel between 75 cities across the U.S., providing expected price movements, confidence levels, historical prices, and buying tips to customers. To further ease airfare shopping, it also allows consumers to shop for airfares for any airport and destination nationwide. Unlike other airfare search engines, it simplifies the shopping experience by showing only the booking options that provide the best value and the lowest fares to consumers. To power its internally developed data mining and predictive analysis software, Farecast.com turned to an enterprise computing grid built using Oracle(R) 10g. The company deployed Oracle Database 10g, Oracle Real Application Clusters, Oracle Partitioning, and Oracle Enterprise Manager 10g on a four-node cluster. Each node is running Linux and is configured with dual Dual-Core AMD Opteron 275HE processors, 8 GB of DDR 3300 ram, and remote NFS attached storage. The migration to an Oracle enterprise grid has enabled Farecast.com to build a fault-tolerant, scalable foundation to support the growing five- terabyte data warehouse the company uses to store and analyze data for its airfare predictions. By using a multi-node cluster, Farecast.com is able to run mixed workloads at a single time, while maintaining high levels of system availability and performance. It's typical for Farecast.com to assign one node of its four-node clustered database to support ad hoc queries, another node to handle data loading, and a third node to support administrative tasks. Company Benefits from Improved Data and System Management Capabilities Farecast.com's decision to migrate to Oracle 10g stretches beyond the redundancy and scalability that Oracle provides. Farecast.com's reliance on Oracle Partitioning has had a significant impact on its ability to more efficiently manage large and growing data volumes. With some of its database tables spanning tens of billions of rows, Farecast.com uses Oracle Partitioning for fast data loads and improved query performance. Oracle Partitioning enhances the manageability, performance, and availability of a wide variety of applications by allowing tables and indexes to be subdivided into smaller pieces, enabling these database objects to be managed and accessed at a finer level of granularity. Oracle Database 10g's data compression capabilities have also had a positive impact on the performance of Farecast.com's data warehouse and related storage capacity. The company has benefited from faster query performance on larger data sets as well as savings on storage capacity with a five to one compression ratio -- meaning that it stores five terabytes of data onto one terabyte of disk. With a single IT manager on staff, the company relies on Oracle Enterprise Manager 10g for systems management and monitoring without the need for additional resources or third-party software. Farecast.com uses Oracle Enterprise Manager 10g for managing backups, performance monitoring, and database tuning, among other tasks, to ensure customers a high quality of service. "The complexity of the data we're looking at really lends itself to one solution and that's Oracle," said Jay Bartot, VP of Engineering, Farecast, Inc. "There are a lot of solutions out there but they don't offer the depth of features or scalability that Oracle has. The ability to run advanced partitioning, compression, querying and management capabilities on a grid of low-cost, industry standard hardware makes it a compelling solution."