Coding techniques for data-storage systems

摘要: As information-bearing objects, data-storage systems are natural consumers of information-theoretic ideas. For many issues in systems, the best trade-off between cost, performance and reliability, passes through application error-correcting codes. Error-correcting codes that specialized for is subject studied by this thesis. On practical side, central challenges storage addressed, both at individual-device level higher enterprise disk arrays. The results individual devices include a new coding paradigm Multi-Level Flash benefits density access speed, also higher-throughput algorithm decoding Reed-Solomon with large radii. arrays address models constructions to combat correlated device failures, introduce highly-regular array-code optimal redundancy updates. theoretical research stretches across multiple layers theory innovation: error models, existing techniques known To bridge properties constraints mathematical language theory, well-motivated abstractions proposed. Among them t asymmetric limited-magnitude errors clustered erasures. Later, after maximizing theory's power addressing abstractions, employ schemes analytically validated.