Description:
As a developer who has wrestled with storing images and other binary data in MySQL, I've experienced firsthand the pain of the current disconnect between data management and storage. The common practice of storing files in the filesystem and referencing them in the database creates a fragmented workflow, leading to increased complexity and potential inconsistencies. This approach forces developers to manage file system interactions separately from database operations, which contradicts the very reason we choose MySQL: its robust data management capabilities.
This experience led me to explore a more integrated approach, where MySQL treats the filesystem as a core component of its data management strategy, not just a storage repository. This is not simply about storing files within MySQL, but about fundamentally rethinking how MySQL interacts with the underlying storage system to unlock significant performance and flexibility gains.
The core idea driving this proposal is to address the real pain of fragmented data management for real gain, especially with the rise of AI applications. Specifically, a custom filesystem designed for MySQL could:
* Simplify Developer Workflow: By integrating storage and management, MySQL can streamline the handling of diverse data types, including binary data and complex AI model structures, reducing the burden on developers.
* Enable Hardware Acceleration for AI Models: A filesystem that facilitates integrity between hardware RAM and GPU storage, combined with hardware-accelerated symbolic links, could unlock the potential for building hardware-accelerated AI models directly within MySQL. This would significantly improve AI model performance and efficiency.
* Position MySQL as a Leader in AI Databases: By providing a robust and efficient platform for AI model development, MySQL can attract significant interest and potential collaboration from AI-focused companies.
This proposal builds upon my previous verified feature requests, which introduce concepts relevant to this integrated approach:
* Bug #117899: Type-Hinted Dynamic Columns (Verified)
* Bug #117910: InnoDB v2: Hardware-Accelerated Relational Data via Symbolic Links (Verified)
These previous requests lay the groundwork for a more unified data management and storage paradigm.
I believe that this strategic shift will not only address a critical developer pain point but also open up the potential for building hardware-accelerated AI models stored directly in MySQL. This could lead to significant community involvement and potential hardware adjustments to fully realize the benefits of this integrated approach.
I am eager to discuss these ideas further and collaborate with the MySQL team to develop a roadmap for implementation.
How to repeat:
this is not a bug. this is the next generation of AI databases.
Suggested fix:
The core of this proposal lies in the development of a custom filesystem optimized for MySQL. This filesystem would be designed to seamlessly integrate data management and storage, enabling efficient handling of binary data and complex AI model structures.
Key aspects of this filesystem would include:
* **Hardware-accelerated symbolic links:** Leveraging the capabilities of modern hardware to optimize relationship traversal.
* **Integrated RAM and GPU storage:** Facilitating efficient data transfer between RAM and GPU memory for AI model training and inference.
* **Granular access control:** Ensuring data security and integrity.
By implementing this filesystem, MySQL can position itself as a leading platform for AI model development, attracting significant industry collaboration and potential hardware adjustments.
