针对现有存储系统不能很好的动态适应工作负载变化情况,本文提出了逻辑进化存储系统的概念,由此设计了一种适合不同磁盘阵列(RAID)之间动态转换的数据分布策略.根据这种策略推导出RAID在同步工作负载情况下的平均响应时间公式,计算出不同RAID级别在各种工作负载情况下的转换门限.为了提供最佳的服务特性,本文还提出根据数据访问的特征自动调节条带单元大小(SUS:Stripe Unit Size)的策略,并推导出磁盘阵列在不同工作负载情况下的最优SUS.最后,通过大量的实验验证了本文的分析结果.
In order to change the status that storage system can not adapt to dynamic workloads,The paper introduced a new concept of logistic evolution storage system and design a data layout which can make transformation among varies RAID levels easy. By analyzing action of RAID1 and RAID5 under synchronous workload, we deducing a approximation analytical expres- sion to predict disk array mean response time and compute threshold of parameters for RAID organization transformation. In order to get the best performance for varies RAID levels,we also introduce a method which adjusts stripe unit size for adapting varies workloads and deduce the optimal stripe unite size. Finally, we carry out a large number of simulations to validate our analytical result.