This is no small task, however. Each of Blizzard’s games have features that the others do not; gameplay in the real-time strategy StarCraft, for example, is completely and fundamentally different from that of the action RPG Diablo. In this presentation, we will describe what our creative teams need in order to make the design decisions that drive our different titles and how those needs are unique to the game and more often than not, even unique to specific game mechanics or features. Data sets even within a single game can be very different; some are high-level business metrics while others are gameplay details that are kept at the finest possible granularity; some are important for a single patch cycle only while others are significant over the entire life of a game.
We will then talk about why classic data warehousing approaches do not meet our needs. We can’t tolerate the loss of agility and responsiveness that those approaches impose due to architectural rigidity and resource bottlenecks. We also need to handle massive amounts of real-time streaming game data without being bound by pre-defined aggregations.
Finally, we will describe how several Hadoop-based technologies have allowed us to deal with more data at a deeper level than ever before. Our horizontally scalable infrastructure allows data to land and be immediately available to our data analysts with little or no development time. Even better, we can keep the data around in full detail so that we can come back later to answer the questions our creative teams didn’t yet know they had.
The presentation will emphasize the critical role users’ needs play in the design, implementation, and adoption of Hadoop within a fast-moving, dynamic environment that supports three Triple A game titles and all supporting business units. Participants are expected to see firsthand how understanding users’ needs can lead to a well-scoped and ultimately successful architectural design. Examples will cover both workflow- and data-design; code samples will feature the use of the Hadoop Streaming API and Python. Examples will feature interesting data sets from Blizzard titles.
Amanda Gerdes is a data engineer with Blizzard Entertainment, focusing on the Blizzard Data Warehouse and its many supporting data pipelines. With eight years of experience in ETL development and management, she currently rounds out a team responsible for providing fast, accurate data at scale to Blizzard’s Business Intelligence team. Amanda holds a B.A. from UC Berkeley as well as an MBA and M.S. in Systems Engineering from Loyola Marymount University. Whenever she catches herself thinking that maybe she’s not doing too badly at this “life” thing, she looks up her World of Warcraft playtime in the data warehouse and is humbled once again.
Brian Griffith’s career spans more than twelve years in the software, financial, and entertainment industries. Prior to Blizzard, he was the lead DBA and data warehouse engineer for Eastern Bank, integrating disparate systems into a single, secure enterprise data system. Currently at Blizzard, he works passionately with vast amounts of data to help game designers make their games even more epic. He holds a B.S. and an M.S. from Northeastern University, specializing in neuroanatomy and statistics. His Blood Elf paladin wears pink armor.
For exhibition and sponsorship opportunities, contact Susan Stewart at firstname.lastname@example.org
For information on trade opportunities with O'Reilly conferences email mediapartners
For media-related inquiries, contact Maureen Jennings at email@example.com
View a complete list of Strata + Hadoop World 2013 contacts