Skip to content

Latest commit

 

History

History
35 lines (22 loc) · 2.88 KB

README.md

File metadata and controls

35 lines (22 loc) · 2.88 KB

Surge

Build Main codecov

WARNING Surge is currently in a pre-release state. There are a lot of great ideas and solid components in this repository, but we are still actively working and shaping the libraries as we move towards a more stable and official 1.0 release. For more details, see what you can currently expect for support and stability.

Many stateful message driven applications today leverage Kafka as a message bus, but tend to keep state stored in a data store outside of Kafka. Keeping this separate data store adds a bit of extra complexity, maintenance overhead, and in bad fault conditions, synchronization issues between the message bus and state store. Surge aims to simplify this by moving the state store to Kafka - providing an engine for stateful streaming and CQRS applications to operate purely on Kafka.

Why the name? We know there are a lot of web frameworks for the JVM that already do a fantastic job at that. Instead of trying to turn Surge into a framework that does it all, we're focused on making Surge a library that easily "plugs in" to other web frameworks in order to provide persistence.

Why Surge?

By leveraging Kafka exclusively for maintaining state, we eliminate the additional overhead of an additional data store exclusively for maintaining state.

Removing the additional data store from the equation, services can lean on async Kafka replication for Disaster Recovery - taking your RPO and RTO from the hours a traditional backup may take to the few seconds it might take to replicate a message in real time over Kafka. Additionally, removing this additional data store can eliminate problems that may occur when the data store itself is lost and must be restored. In these cases, your service may have published some events that are now not reflected in your restored data store - leading to confusion and at worst corrupt data leaving your service onto the event stream.

By leveraging Surge, you get a highly scalable, resilient, reactive mechanism for state management purely on Kafka that simply avoids all of these potential problems by ditching traditional state storage in favor of bringing it to where the rest of your data lives anyways - Kafka.

Reference Documentation

Our current documentation can be found at https://ultimatesoftware.github.io/surge/.

Contributing

Contributions are welcome!

Check out our contributing guidelines for more details about the process. Feel free to additionally ask for clarifications or guidance in GitHub issues directly.