LDPC code design for distributed storage systems
Loading...
Date
2023
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Thesis (Ph.D.) - Bogazici University. Institute for Graduate Studies in Science and Engineering, 2023.
Abstract
An exhaustive study on a cooperative repair scheme for distributed storage systems (DSSs) utilizing base station (BS) assistance was presented, involving the concept of an admissible region and optimization problem formulation. Findings revealed a broader admissible region, optimizing the storage size-repair bandwidth cost tradeoff. Moreover, a dynamic DSS environment model was introduced. The numerical analysis highlighted the superiority of the BS-assisted cooperative scheme over traditional approaches. Furthermore, an in-depth analysis was made concerning the interplay among the decoding threshold, average repair bandwidth, and code rate, crucial for optimizing low-density parity-check (LDPC) code design for a DSS. The influences of two proposed repair protocols - the random access repair protocol and the ideal repair protocol - on these relationships were emphasized. For random access repair protocol, a method for determining the check node degree distribution ρ(x) to achieve the minimum average repair bandwidth was proposed. Using numerical analysis, ideal repair protocol’s superiority was demonstrated. Moreover, the theoretical findings were verified, emphasizing the potential for further repair bandwidth improvements. Lastly, the identification and elimination of stopping sets in a DSS with LDPC-encoded data were explored. Additionally, a greedy algorithm for determining the repair bandwidth required for a failed storage unit was suggested, indicating potential reductions in the bandwidth needed for single-node repair.