Replicated state machines are the cornerstone of distributed systems - and at the heart of a replicated state machine is a transactional log. While these logs are fundamental to replication in a distributed system, they have recently emerged as the glue for event driven systems. In order processing of logs enables multiple developers to build a diverse set of applications on top of the same stream ranging from materialized views, event driven systems, realtime analytics, etc. This powerful primitive has given rise to log streaming systems like Kafka, Storm, Amazon Kinesis, Amazon Lambda, and DynamoDB Streams, among others.

This talk will provide a deep dive into ordered logs as a primitive, patterns in log based systems, and the capabilities that they make possible. You will get a perspective from two of the engineers behind DynamoDB Streams on what it takes to build an ordered, highly available, durable, performant, and scalable replicated log stream. You will walk out with a passion to build an event driven system - with an understanding of how event processing systems work and how you can use them to simplify development of a complex distributed system.