Want to roll out a database active active over multiple datacenters? Want to be able to do rolling upgrades and software releases without downtime? How about scale to millions of durable writes a second with single digit millisecond response times.;These requirements really cut down your database choices, and any database you’re considering are probably heavily influenced by Amazon’s 2007 paper: Dynamo: Amazon’s Highly Available Key-value Store. ;In this talk we’ll go through the main parts of Dynamo and discuss them in the context of how Apache Cassandra has implemented them. We’ll cover:;+ Consistent hashing;+ Strategies for replication;+ Working out distributed state with Gossip;+ Recovering from failure with Hinted handoff and Anti-entropy repair;We’ll also talk about why picking a masterless architecture, and trading availability and partition tolerance for consistency really can lead to a linearly scalable database.