A multithreaded fork of Redis that is faster (docs.keydb.dev)

Article summary

KeyDB is a multithreaded fork of Redis that claims to be 5x faster with nearly 5x lower latency. The project aims to provide a more powerful and simplified alternative to Redis, with features like Active Replication and multithreading. KeyDB's architecture allows for concurrent execution of network I/O and query parsing, while maintaining atomicity guarantees through locking. The project provides benchmarking data and instructions on how to reproduce the tests.

Main themes

• Redis fork
• multithreading
• performance optimization
• database architecture
• compatibility and simplicity
• benchmarking and testing
• production environments and use cases

What commenters say

• Some users are skeptical about the need for a multithreaded Redis fork, citing the single-threaded nature of Redis as a benefit for simplicity and atomicity.
• Others argue that multithreading can significantly improve performance in certain scenarios, such as high-traffic environments or complex workloads.
• The use of a binary client-server protocol could potentially improve performance by reducing parsing overhead, but it may require significant changes to the Redis codebase and client libraries.
• Some commenters have expressed interest in exploring the use of KeyDB in production environments, while others have reported mixed results in testing the software.
• There is disagreement about the effectiveness of running multiple instances of Redis on the same machine as a way to achieve parallelism, with some arguing that it can lead to data duplication and wasted resources.
• The importance of maintaining compatibility with Redis is highlighted, with some commenters noting that KeyDB's goal of 100% compatibility may limit its ability to innovate and improve performance.
• The trade-offs between simplicity, performance, and maintainability are a recurring theme in the discussion, with different commenters prioritizing these factors differently.