This document discusses Hadoop usage at eBay over time from 2007 to 2015. It describes:
- The growth of eBay's Hadoop clusters from 1-10 nodes in 2007 to over 10,000 nodes and 150,000 cores projected for 2015.
- How the amount of data stored in Hadoop has grown from 1PB in 2010 to a projected 150+ PB in 2015.
- The types of clusters eBay uses including dedicated, shared, and HAAS clusters.
- Some key use cases for Hadoop at eBay like building a near real-time search index and processing 1.68 million items in 3 minutes.
- Operational requirements for eBay's large Hadoop ecosystem like high availability, security,