Dear Apache James Development Community, At Linagora, we have encountered challenges related to Redis's memory usage when employed as a storage backend for Rspamd's Bayesian learning. Redis, being an in-memory data store, can experience significant memory consumption as the volume of statistical data grows, quickly to reach the costly memory limit. As a result, we are exploring the potential of integrating Apache Kvrocks as a potential alternative for Redis within the Apache James server. Our primary motivation stems from the need for scalable storage solutions, particularly for handling Rspamd Bayes learning data.
We believe that integrating Kvrocks could offer several advantages, including: - Enhanced scalability of storage capabilities. - Seamless compatibility with existing Redis protocols, facilitating a smoother transition. - Friendly Apache license and ecosystem. Our team is eager to experiment with this integration and contribute to its development. We plan to document the setup process, share our findings, and collaborate closely with the Apache James community about the topic. To start this exploration, we have submitted a pull request to the Apache James project, aiming to experiment Apache Kvrocks with the James test suite at: https://github.com/apache/james-project/pull/2660. We have initiated discussions with the Apache Kvrocks community to assess the feasibility and benefits of this integration. You can find our correspondence with the Kvrocks developers here: https://lists.apache.org/thread/f6n8qpl2fnho472b63o7r4d7ygsqz3t4 We invite feedback, suggestions, and any insights you might have regarding this initiative. Together, we can explore the possibilities of this integration and work towards enriching the Apache James ecosystem. Best regards, Quan
