The Apache Cassandra database is the right choice when you need scalability and high availability without compromising performance.
That may very well be true, but it took us a year to get there and that road was dotted with site’s lowered availability and compromised performance.
In this talk I am going to share the insights we gained when we migrated parts of the Yammer’s messaging pipeline from a custom storage solution backed by Java BDB to Cassandra. Covering topics like:
- Why we decided to move to Cassandra from our proprietary Berkley DB backed database
- Modeling data and capacity supported by metrics
- Zero downtime production rollout
- How things started falling apart after three months of seamless operation
- How to diagnose and fix things later on
Yammer is a private social network that helps employees collaborate across departments, locations and business apps. We currently have 10 million users and are planning to grow at least 10 times. To achieve that, our systems needs to be cheap and easy to scale and resilient to failure.