Reed-Solomon MDS Coding with Composite Fields for Faster Decoding
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing Reed-Solomon code constructions for error correction and erasure correction are computationally costly due to finite field operations, especially as the number of symbols increases, and recent improvements have not adequately addressed performance for moderate k in communication and storage systems.
Innovation Solution
The use of composite extension finite fields and a deterministic algorithm to maximize the number of columns in the generator matrix with elements in smaller finite fields, allowing for efficient encoding and decoding operations, particularly leveraging SIMD capabilities for performance gains.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional Reed-Solomon code constructions are used, then error correction capability is maintained, but computational cost increases significantly as the number of symbols increases
Solution Approach 1:
The patent segments the finite field operations by decomposing elements of GF(2^s) into vectors of elements from GF(2^r) where r divides s. This segmentation allows the generator matrix to be constructed with columns containing elements from smaller subfields, reducing the computational complexity of finite field operations while maintaining the error correction capability of Reed-Solomon codes.
Solution Approach 2:
The patent changes the parameter representation by expressing field elements in terms of their coordinates with respect to a basis of the extension field. By representing GF(2^s) elements as vectors over GF(2^r), the patent transforms the computational parameters to enable more efficient arithmetic operations, particularly for encoding and decoding processes.
2Reliability
If generator matrices with elements from larger finite fields are used, then code performance is optimized, but encoding and decoding speed decreases
Solution Approach 1:
The patent segments the generator matrix columns to maximize the number of columns containing elements from the base field GF(2) or smaller extension fields GF(2^r). This segmentation strategy allows the system to maintain optimized code performance while significantly improving encoding and decoding speeds by reducing the computational burden of large finite field operations.
Solution Approach 2:
The patent applies local quality by allowing different columns of the generator matrix to have elements from different subfields. Specifically, columns are chosen to contain elements from the smallest possible subfields (GF(2), GF(2^r1), GF(2^r2), etc.), creating local variations in field complexity that optimize both performance and speed for different portions of the code.
3Measurement precision
If complex finite field operations are performed, then accurate error correction is achieved, but processing time increases
Solution Approach 1:
The patent introduces intermediate representations of field elements as vectors over smaller subfields. This intermediary representation allows accurate error correction to be achieved through a series of simpler operations on vector components rather than direct complex finite field arithmetic, thereby reducing processing time while maintaining correction accuracy.
Solution Approach 2:
The patent changes the dimensional perspective by representing single field elements from GF(2^s) as multi-dimensional vectors over GF(2^r). This dimensional transformation enables error correction operations to be performed component-wise on the vector representations, reducing the time complexity of operations while preserving the mathematical properties needed for accurate error correction.
Data Source
AI summary
Disclosed apparatus and method improve the computational efficiency of encoding and decoding data having erasures according to a maximum distance separable (MDS) code based on a Reed-Solomon code. Thus, n encoded fragments are formed by multiplying k data fragments by an n×k generator matrix for the MDS code. The code is formed by reducing, in the generator matrix to the extent possible, the size of the finite field to which entries belong—in some cases to the base field having only two elements. In this way, unlike codes known in the art, the generator matrix has more than one column whose entries each take values in the finite field having two elements. In some cases, the generator matrix has a column whose entries each take values in one or more intermediate fields between the finite field having two elements and the encoding field.