AWS re:Invent 2025-Disagree in Commits:The Performance Improvements That Cut Costs by a Third-OPN309

AWS re:Invent 2025 - Disagree in Commits: The Performance Improvements That Cut Costs by a Third

Overview

• This presentation discusses the journey of the Amazon Elasticache team in contributing to the open-source Reddis project and the subsequent creation of the Valky open-source fork.
• The key focus is on how the team's open-source contributions led to significant performance improvements and cost savings for the Elasticache service.

Elasticache's Open-Source Journey

• Initially, the Elasticache team had a private fork of Reddis to add features like TLS support, which was challenging to maintain due to merge conflicts.
• The team decided to focus on contributing bug fixes and performance optimizations upstream to the Reddis project.
• This process was initially difficult due to the approval process required to contribute to open-source at Amazon.
• However, the team persisted and was able to contribute a fix for a bug related to password handling in Reddis replication, which was well-received by the Reddis maintainers.

Valky: The Reddis Fork

• When Reddis changed its licensing to SSPL, which was not recognized as open-source by the OSI, the community forked Reddis 7.2.4 to create Valky under the BSD license.
• Valky was created under the governance of the Linux Foundation, with a technical steering committee that included the original Reddis maintainers and representatives from various companies.
• The Elasticache team joined the Valky project and continued to contribute performance improvements and features.

Technical Innovations in Valky

• Horizontal and vertical throughput improvements: Valky was able to scale from 0 to 5 million requests per second in 15 minutes, compared to 70 minutes for Reddis 7.2.
• Memory efficiency improvements: Ericsson contributed a memory-efficient hash table that resulted in a 41% memory savings for a large customer.
• Vector similarity search: Valky incorporated a better-performing vector similarity search implementation contributed by Google Cloud.

Business Impact

• The performance and efficiency improvements in Valky have led to significant cost savings for Elasticache customers:
  • Serverless deployments are 33% less expensive than the Reddis equivalent.
  • Node-based deployments are 20% less expensive than the Reddis equivalent.
• Reddis has been pulling in many of the performance improvements from Valky, demonstrating the benefits of open-source collaboration.

Conclusion

• The Elasticache team's journey showcases the value of open-source collaboration and how it can drive innovation, performance improvements, and cost savings for both the service provider and its customers.
• The creation of Valky and the ongoing contributions to the open-source community have been a win-win for all involved.