A Heterogeneous Hybrid Storage Method Based on Ceph Erasure Code
Mingyuan Qin, Zheng Fu · 2021
In order to achieve the goal of reducing storage redundancy, Ceph provides support for erasure code technology, which can more effectively use storage space. However, when Ceph selects storage nodes, it only uses storage capacity as the only indicator, without considering the resource load status and the heterogeneous situation of nodes, which makes a large number of read and write tasks allocated to storage nodes with lower performance. The above causes the problem of low reading and writing efficiency in Ceph erasure code mode. Based on this problem, we propose a heterogeneous hybrid storage method based on Ceph erasure code. Under the premise of comprehensively considering node CPU utilization, memory utilization and other performance indicators, Using TOPSIS method to implement multi-attribute decision-making and assigning Ceph erasure code tasks to the storage node with best performance. Experiments show that the method can effectively improve the read and write performance of large objects in Ceph erasure code mode.