July 3rd, 2013


This session by Jay Patel will cover various use cases for Cassandra at eBay. Jay starts with an overview of eBay’s heterogeneous data platform comprised of SQL & NoSQL databases, and where Cassandra fits into that. For each use case, Jay will go into detail of system design, data model & multi-datacenter deployment. To conclude, Jay will summarize the best practices that guide Cassandra utilization at eBay.


eBay Statistics

• Over 100 Million Users, Sellers and Buyers worldwide

• $75 Billion worth of goods sold on eBay per year

• Billions of page views per day

• Petabytes of Data

• Always available, always online


eBay Cassandra Usage

• 10+ Clusters, 100+ Nodes

• 250 TB Provisioned (Local HDD + Shared SSD)

• 9 Billion Writes per day

• 5 Billion Reads per day


Take Aways by Jay Patel

• One Size Does Not Fit All

         – Use Cassandra for the right use case

• Choose Proper Replication Factor and Consistency Level

         – They offer latency, availability, durability, consistency and cost

         – Cassandra supports tunable consistency, but remember strong consistency is not free

• Many Ways to Model Your Data in Cassandra

         – The best way depends on your use case and query patterns

• De-normalize and duplicate for read performance 

         – But don’t renormalize if you don’t need to






If you’re looking for additional information on data modeling, check out Jay Patel’s slides ‘Cassandra Modeling — Best Practices and Examples