面向内存云的协调器选举策略

发布时间：2018-02-24 04:22

本文关键词： 内存云协调器单点失效适应度函数选举策略　出处：《计算机应用》2016年09期 　论文类型：期刊论文

【摘要】：针对Zoo Keeper机制难以满足内存云(RAMCloud)低延迟、快恢复的问题,提出了一种面向内存云的协调器选举策略(CES)。首先根据内存云网络环境与协调器自身因素将协调器性能指标分为个体指标与协调器间指标两类并分别建立模型;然后将内存云的运行分为正常运行期与数据恢复期两阶段并分别建立适应度函数,再按时间比合并为总适应度函数;最后在备选协调器(RBC)的适应度值的基础上提出一个具备稳定择优性与随机性的新算子,CES首先通过筛选来排除性能较差的个体,缩小选择范围后再在理想协调器的集合中采用轮盘赌方法选择最终的个体。实验结果表明,在NS2仿真环境下CES选择的个体相比其他备选协调器数据处理延迟降低了19.35%;在搭建的内存云环境中,与Zoo Keeper机制相比,CES的选择结果在快速恢复中时间减少了10.02%。在内存云的实际应用中,CES在处理单点失效问题上能有效选择性能更优的协调器,确保了低延迟、快恢复的要求。
[Abstract]:According to the Zoo Keeper mechanism to meet the cloud memory (RAMCloud) low latency, fast recovery problem, proposes a memory oriented cloud coordinator election strategy (CES). According to the cloud memory network environment and the coordinator will own factor coordinator performance index is divided into individual index and coordination between indexes of two types respectively. The establishment of model; and then the memory of cloud operation recovery period the two stage is the normal operation period and the data were established according to the fitness function, time is combined for a total fitness function; finally, in the alternative Coordinator (RBC) of the base to adapt the value to put forward a new operator with optimality and stochastic stability the first CES by screening to exclude the poor performance of the individual, reduce the selection range and then set in the ideal coordinator in the final choice of individual roulette method. The experimental results show that CES in NS2 simulation environment Choose the individual compared to other alternative coordinator data processing delay is reduced by 19.35%; in memory cloud built environment, compared with the Zoo Keeper mechanism, the choice of CES results in a decrease in 10.02%. in the practical application of memory in the cloud in the fast recovery time of CES in the treatment of single point failure problem can effectively select performance coordinator better, to ensure low latency, fast recovery requirements.

【作者单位】：新疆大学软件学院;新疆大学信息科学与工程学院;
【基金】：国家自然科学基金资助项目(61262088,61363083,61462079,61562086)~~
【分类号】：TP333

【相似文献】