Strange Loop

Apache Cassandra: Pragmatic Consistency for Massive Scale

A recent burst of interest in alternative databases has been highlighted by the creation of the cleverly (mis)named NoSQL group. But the reasons why someone might want to in some cases give up their relational databases are still very misunderstood. One of the primary reasons is scalability.


In the world of relational databases, consistency is typically strong and immediate, but (as explained by the CAP theorem) large datasets can gain other advantages by allowing for eventual consistency, including better read/write performance and high availability, even in the face of multiple machine failures. Apache Cassandra is a third-generation distributed database providing BASE (rather than ACID), allowing it to scale linearly and replicate elastically across multiple datacenters, as demonstrated by its creators at Facebook.


This talk will briefly cover the advantages of eventually consistent, non-relational databases like Apache Cassandra, but will primarily use Cassandra's API and data model to explore common use cases of strongly consistent storage which can easily be modified to support eventual consistency.

Stu Hood

Stu Hood is a contributor for the incubating Apache Cassandra project, and Technical Lead for the Search team in the Email & Apps division of Rackspace. He enjoys interrupting conversations to help solve problems, and building scalable systems using open source components. His case study about Rackspace's usage of Hadoop was recently published in Tom White's O'Reilly book "Hadoop: The Definitive Guide".