Efficient Metadata Management in Large Distributed Storage Systems

Efficient metadata management is a critical aspect of overall system performance in large distributed storage systems. Directory subtree partitioning and pure hashing are two common techniques used for managing metadata in such systems, but both suffer from bottlenecks at very high concurrent access rates.We present called Lazy Hybrid (LH) metadata management that combines the best aspects of these two approaches while avoiding their shortcomings.