云存储的多维离线调度算法

摘要

由于使用云块存储系统是OpenStack的Cinder,且Cinder默认使用的调度算法只是从存储节点的可用空间容量方面考虑,所以它不能够有效地保障卷请求的IOPS.针对这种情况,综合考虑存储节点的可用空间容量和I/O吞吐量两个方面,同时基于APX-hard 多维向量集装箱(Vector Bin Packing)算法,提出了一个多维离线调度算法(MVCFD).实验结果表明,该算法与现有的块存储调度算法相比,在卷请求的空间与IOPS两方面都满足的情况下,还能够有效地减少活动主机的数量.%Since the cloud block storage system used is the Cinder of OpenStack, and the default scheduling algorithm used by Cinder is only from the available space capacity of the storage node, it can not guarantee the IOPS of the volume request effectively.In view of this situation, considering the available space capacity and I / O throughput of the storage node, a multidimensional off-line scheduling algorithm (MVCFD) is proposed based on the APX-hard Vector Bin Packing algorithm.Experimental results show that the proposed algorithm can effectively reduce the number of active hosts in the case of both the space request and the IOPS of the existing block storage scheduling algorithm.