为满足纠删码更新方法的效率需求, 提出一种基于协作方式的更新方法Coop-U, 以批量形式高效完成纠删码数据的多点更新.通过一种三层的更新框架保证数据更新的一致性, 通过一种负载感知的分组方法动态调整分组大小, 通过一种协作式的数据处理算法有效组织多层之间数据的传输与计算.基于数据驱动的模拟器对比实验结果表明, 与典型的纠删码数据更新方法 Data-R和Full-W相比, Coop-U的平均更新时间分别减少了42％和30％.%To meet the efficiency requirement of the update schemes of erasure codes, a cooperative update scheme for multiple updates with erasure codes was proposed, called Coop-U.The data block was updated eagerly and the parity blocks were cooperatively updated in bulk.A three-layer update structure was proposed to ensure the generality.For efficiency, a workload-aware grouping algorithm was proposed to group the nodes to be updated and the group size was dynamically adjusted according to the update workload.A cooperative data processing algorithm was proposed to organize the data transmission and computation among layers.To evaluate the performance of Coop-U, it was implemented on HDFS-RAID and testbed experiments were conducted on different update schemes.Extensive experiments verify that Coop-U reduces the update time by 42％ and 30％ on average compared with the typical update schemes Data-R and Full-W respectively.
展开▼
