Compressive sensing-enhanced distributed computing algorithm for large-scale data storage optimization

Abstract

The continuous growth of large-scale distributed data systems has raised critical challenges in achieving efficient storage, fast reconstruction, and balanced resource utilization under heterogeneous environments. Traditional methods often suffer from high redundancy, uneven load distribution, and degraded scalability. To address these issues, this paper proposes a novel Compressive Sensing-Enhanced Distributed Optimization Framework (CS-DistOpt) that unifies compressive sensing, consensus optimization, and distributed computation for large-scale data storage. The proposed framework dynamically adjusts compression ratios according to node capability and network condition, employs geometry-aware sensing matrices to improve reconstruction stability, and integrates consensus-based recovery to reduce global error variance. A series of experiments conducted on a 32-node cluster demonstrate that CS-DistOpt achieves up to 31% improvement in throughput and 25% reduction in reconstruction error compared with baseline methods. Furthermore, system-level optimizations such as sparse-operator fusion and mixed-precision quantization further enhance computational efficiency without sacrificing fidelity. These results confirm the effectiveness of combining compressive sensing with distributed optimization, providing a scalable and reliable foundation for future high-performance data storage infrastructures.