Case Study - Adgooroo
Configuration: EX-5000 with BobCat Connector
Data: Web-scale SQLServer database, 100,000,000+ records
Adgooroo provides its customers with real time access to web advertising data that allows companies to get an edge over their competitors by providing detailed reporting on trends in on-line advertising. Each day, Adgooroo samples millions of ads and keywords by probing thousands of online advertising outlets, and recording the ads and keywords. This massive amount of data is inserted into Microsoft SQLServer databases for further analysis and batch processing. With the explosive growth in online advertising, Adgooroo has an extreme challenge in managing this volume of web-scale data. To meet the challenge, Adgooroo turned to MaxxCAT for extreme performance. "Our overnight indexing process was taking almost 24 hours to complete on a cluster of fully configured high end multi-core Dell servers. The MaxxCAT appliance is doing a full text index of this same data set in under 4 hours on a single node. This is on a database with 100 million rows and growing" reports Adgooroo. "Query response time is amazing, we are getting sub-second responses when querying this very large database and it has enabled us to build a much more usable system".
The Adgooroo implementation is an example of how MaxxCAT appliances go beyond user-level search and can be embedded as compute servers for other applications. With the flexible BobCat database connectors, MaxxCAT can quickly extract data from host databases and because of the innate indexing and query performance of the machines, the appliances can serve as a specialized processor or a "hardware subroutine", providing a simple API to the complex problem of index and retrieval. Because MaxxCAT appliances are designed from the ground up to do only one thing, search, they do it extremely well. They do it much better than general purpose hardware, or specialized software running on general purpose hardware because the entire machine from the low level disk drivers and memory management are optimized for full text search and nothing else.