A Coding Method for Node Restoration in Distributed Storage System

Abstract

The invention discloses a coding method for repairing a node of a distributed storage system. Coding parameters in the distributed storage system are extracted; a coding model of the storage system is established; a protection column and a design column are introduced, thereby obtaining an expanded coding model; a proportion between the protection column and the design column is determined according to a partial derivative of generalized reparation bandwidth, thereby obtaining the values of the protection column and design column; and establishing an additional information construction matrix according to the values of the protection column and design column, constructing additional information, and superimposing the additional information to check information of the design column, thereby finishing code construction of generalized piggybacking codes. According to the method, no extra storage cost exists; in the node reparation process, the original decoding feature of the storage system can be kept unchanged; moreover, no complex matrix operation is carried out in the node reparation process; only simple linear superposition is carried out; the complexity is low; and the method is convenient for realization and has good practical value.

Description

Technical field [0001] The present invention relates to the field of distributed storage, in particular to an encoding method for node repair in a distributed storage system. Background technique [0002] Distributed storage systems are widely adopted by large-scale network applications due to their low investment cost and high storage efficiency. However, due to frequent failures, maintenance and other factors, the reliability of each separate storage node of the distributed storage system has become An important factor affecting the performance of a distributed storage system. At present, error control coding techniques represented by erasure codes have been applied to distributed storage systems. [0003] Maximum distance separable (MDS) code is an erasure coding technique most commonly used in storage systems. In an MDS code with a parameter of (n, k, d), n represents the code length and k is the information The dimension of, its minimum Hamming distance is d=n-k+1. For fixe...

AI Technical Summary

Problems solved by technology

However, there are several big problems in the use of the regeneration code in the actual system: 1. The regeneration code cannot be upgraded on the basis of the original code of the storage system, but the entire system needs to be redesigned and deployed, and the cost of system transformation is relatively high; 2. 1. The regeneration code is not valid for all code parameters (n, k are any integer satisfying n>k); 3. The construction process of the regeneration code is complicated, and the computational complexity of the repair process is also very large, so it is hardly used at present. Distributed Storage System of Regenerated Code

Images