NADE: nodes performance awareness and accurate distance evaluation for degraded read in heterogeneous distributed erasure code-based storage

Abstract

To ensure data availability and save storage space, storage systems usually save data across multiple storage nodes (or servers) using erasure codes. Storage systems need to reconstruct the complete data to respond to reading requests in the case of the loss of some data blocks when node failure occurs. However, a degraded read in erasure code-based storage systems does not fully utilize node resources and ignores the node’s topology. In this paper, we propose a real-time performance evaluation model for storage nodes to evaluate the performance of each node combining a metrics choice and an analytic hierarchy process. We also design a cost evaluation method to calculate the transmission cost by considering the node’s topology. By combining the node evaluation method and a distance calculation, we propose an adaptive degraded read optimization strategy, NADE. We further implement the node selection method NADE in Ceph. The evaluation results show the efficiency of the proposed method.