基于混合策略的云计算弹性扩展设计与实现

发布时间：2018-05-21 15:42

本文选题：云计算 + 弹性扩展　；参考：《大连理工大学》2014年硕士论文

【摘要】：云计算(Cloud Computing)是一种新型的分布式计算范式。它将计算任务分布在大量计算机构成的资源池上,使各种应用能够根据需求获取计算力、存储空间和各种软件服务。云计算用户可以获得其所需的资源,并且按资源使用量付费,就像日常生活中用电一样,使用方便,价格低廉。越来越多的人开始关注和使用云计算,其中一个重要的原因是其弹性扩展特性。云计算平台针对应用负载的动态变化,实时地对应用所分配的资源做出调整,具体的说就是当应用负载增加时,分配给应用更多的资源,防止服务质量Qos下降,保证应用的正常运转；当应用负载减少时,回收相应的空闲资源,增加了资源的利用率,同时也减少了云计算用户的费用。由于应用负载变化的不确定性以及相对规律性,本文提出了一种基于混合策略的云计算弹性扩展解决方案。该方案将神经网络算法与静态规则算法有效地结合在一起,构建起了一个混合的资源需求模型,由此生成弹性扩展规则,解决了单纯基于神经网络模型面对不符合特定模式特征的负载时预测不准确的问题。为了验证本文提出的解决方案的有效性,本文设计了一系列验证实验。实验结果表明：相比于单纯基于神经网络的资源需求模型,本文提出的混合模型的预测准确率更高,获得的弹性扩展规则更加合理。
[Abstract]:Cloud computing is a new distributed computing paradigm. It distributes computing tasks in a large number of computer resource pools, which enables various applications to acquire computing power, store space and various software services according to their requirements. Cloud computing users can get the resources they need and pay for the amount of resources they use, just like electricity in daily life, it is easy to use and low in price. More and more people begin to pay attention to and use cloud computing, one of the important reasons is its elastic expansion. According to the dynamic change of application load, cloud computing platform adjusts the resources allocated to the application in real time. Specifically, when the application load increases, more resources are allocated to the application to prevent the quality of service (Qos) from declining. Ensure the normal operation of the application; when the application load is reduced, the recovery of the corresponding idle resources, increase the utilization of resources, but also reduce the cost of cloud computing users. Due to the uncertainty and relative regularity of application load change, this paper proposes a hybrid strategy based solution for cloud computing elasticity expansion. By combining neural network algorithm with static rule algorithm effectively, a hybrid resource requirement model is constructed, which generates elastic expansion rules. The problem of inaccurate prediction based on the neural network model is solved when the load does not conform to the characteristics of the specific pattern. In order to verify the effectiveness of the proposed solution, this paper designed a series of verification experiments. The experimental results show that the proposed hybrid model has higher prediction accuracy and more reasonable elastic expansion rules than the resource demand model based on neural networks alone.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP393.09

【共引文献】