Preemptive Slice Reading for Dispersed Storage Availability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing data storage systems in dispersed networks face challenges in maintaining data integrity and availability due to storage unit failures and errors, requiring redundant copies and complex error correction mechanisms.
Innovation Solution
A dispersed storage network architecture that uses error encoding techniques, such as Cauchy Reed-Solomon encoding, to distribute data across multiple storage units, allowing for error detection and correction without the need for redundant copies, ensuring data availability and security.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If redundant copies of data are stored to ensure data availability and integrity, then data reliability is improved, but storage cost and device complexity increase
Solution Approach 1:
The patent segments data into multiple encoded slices distributed across different storage units. Instead of storing complete redundant copies, the data is divided and encoded such that any subset of slices can reconstruct the original data, reducing total storage requirements while maintaining availability
Solution Approach 2:
The patent uses error correction coding to create encoded copies of data slices. These encoded copies allow reconstruction of original data without requiring identical redundant storage, achieving data protection with reduced storage overhead
2Reliability
If redundant copies of data are stored to ensure data availability and integrity, then data reliability is improved, but system complexity increases
Solution Approach 1:
The patent changes the mathematical parameters of data representation through error correction encoding. By transforming data into encoded form with specific algebraic properties, the system achieves automatic error detection and correction capabilities without complex procedural mechanisms
3Reliability
If multiple read requests are issued to handle storage unit failures, then data availability is maintained, but access time increases
Solution Approach 1:
The patent performs preliminary error correction encoding during data writing, preparing encoded slices in advance. This preliminary action ensures that when read requests occur, the system can immediately retrieve and reconstruct data without needing to issue multiple sequential read requests, reducing access time
Data Source
AI summary
A read request specifying particular data stored in a set of distributed storage (DS) units is received. The DS units store error encoded data slices in a dispersed manner, and a read threshold number of the slices are needed to recover the particular data. DS units that hold the error encoded data slices are identified, and a number of slice read requests predicted to fail to return a requested slice is estimated. A subset of DS units equal to a read threshold, plus an additional number of DS units determined based, at least in part, on the estimated number of slice read requests predicted to fail, is selected. Read slice requests are transmitted to the subset of DS units, so that even if some of the slice read requests fail, at least a read threshold number should succeed.