Matrix-Based Erasure Coding for Fast Distributed Data Reconstruction

The patent divides the storage system into separate components: file managers that handle logical responsibility for data and storage managers that handle actual data maintenance. This segmentation eliminates single points of failure by distributing responsibilities across multiple specialized components, allowing the system to tolerate component failures while maintaining data availability through the separation of concerns between file managers and storage managers.

The patent introduces file managers as intermediary components between clients and storage managers. File managers handle logical operations and coordinate with storage managers for actual data maintenance, creating a mediator layer that improves reliability by preventing single points of failure while managing the complexity through standardized interfaces and protocols.

The patent segments the error correction and data reconstruction process across multiple storage managers rather than concentrating it in a single component. Each storage manager independently handles maintenance of its assigned data blocks, allowing parallel reconstruction operations that reduce overall reconstruction time while maintaining data integrity through distributed error correction capabilities.

The patent implements preliminary error correction coding during the initial data storage phase, distributing encoded data across multiple storage managers. This preliminary action enables faster reconstruction during failures because the error correction capability is already embedded in the distributed data structure, eliminating the need for time-consuming centralized computation during recovery.

The patent implements distributed hash tables that serve multiple functions: locating data blocks across storage managers, coordinating error correction operations, and managing the mapping between logical data and physical storage. This multi-functional approach improves productivity by enabling efficient data location while managing architectural complexity through a unified distributed coordination mechanism.

The distributed hash table acts as an intermediary layer between file managers and storage managers, providing efficient data location capabilities without requiring complex peer-to-peer communication protocols. This mediator structure improves productivity by enabling fast data retrieval while managing the complexity of distributed coordination through a standardized hash-based lookup mechanism.