Running distributed systems in production can be tremendously challenging. In this session, we will cover common problems and failures seen with distributed systems, and discuss design patterns that can used to maintain data integrity and availability when everything goes wrong. We will use Druid as a real world case study of how these patterns are implemented in an open source technology.

Attendees will learn first hand about the multitude of software, hardware, network, and data center problems that can arise with running distributed systems, and the features that are required for availability and survivability. To provide real world examples, we will examine the architecture of Druid, and how the system is designed to power applications that need to be up 24/7. We will also cover common pitfalls with running distributed systems in various environments, including the trade-offs with on-premise and cloud deployments.

Finally, we will cover best practices around properly instrumenting monitoring and alerting for distributed systems. We will examine various open source technologies that can be used for efficient monitoring, and how these technologies can be used to maintain the availability of your cluster.

We hope this session will help you better use, design, and monitor your data systems.

Additional Links:

• Intro reading list about distributed systems: 

          http://the-paper-trail.org/blog/distributed-systems-theory-for-the-distributed-systems-engineer/

• More information about Druid: www.druid.io