ON-SSD Erasure Coding With Uni-Directional Commands for Lower I/O

The patent segments the centralized parity computation task into distributed operations across multiple SSD member drives. Each SSD performs local parity computations independently using uni-directional commands, eliminating the need for a central controller to gather all data and perform computations. This segmentation reduces data transfer overhead while maintaining data reliability through distributed erasure coding.

The patent extracts the computational burden from the central storage controller and relocates it to the SSD member drives themselves. By implementing erasure coding logic directly in the SSDs, the system removes the bottleneck of centralized computation and the associated data transfer overhead, while still achieving the same data protection goals.

The patent divides the erasure coding functionality into independent segments that can be executed by each SSD individually. This modular approach allows the system to scale horizontally by simply adding more SSDs without increasing the computational burden on a central controller, thereby improving scalability while maintaining data reliability through distributed parity computation.

The patent transitions from a centralized vertical architecture to a distributed horizontal architecture. By moving erasure coding operations from the central controller dimension to the distributed SSD dimension, the system achieves better scalability as each SSD operates independently, allowing linear scaling with the number of drives.

The patent inverts the traditional command direction by using uni-directional commands instead of bi-directional commands. Instead of the controller sending commands and receiving responses for each EC operation, the SSDs autonomously execute erasure coding operations based on simple write commands, significantly reducing the number of I/O operations and protocol complexity while maintaining data integrity.

The patent enables SSDs to perform erasure coding operations autonomously without requiring continuous controller intervention. Each SSD independently computes and stores parity information using local resources, reducing the I/O overhead associated with bi-directional communication while ensuring data reliability through self-contained error correction capabilities.