移动云计算中弹性划分方法研究

发布时间：2019-01-11 11:50

【摘要】：近些年来,移动应用程序和移动设备都取得了巨大的发展。但是与PC相比,移动设备依然无法运行数据密集型应用程序。这些设备都会受到电池电量、屏幕大小、无线通信等限制。为了突破这些限制,我们可以采用移动云计算技术。移动云计算可以定义为云计算的延伸。云计算具有强大的计算能力、海量的数据存储功能以及提供新的商业模式,即计算资源以服务的方式提供给用户。移动云计算的硬件资源还包含了移动设备。这样就可以利用不同移动设备的集体传感、网络连接、存储以及计算能力来建立一个分布式的基础架构。将移动应用程序迁移到移动云平台会面临弹性划分问题。弹性划分问题是一项复杂的问题。弹性划分的结果会受到诸多因素的影响,其中包括输入因素,限制因素以及划分目标。输入因素与电池电量,网络状况,设备负载,云负载以及当前执行性能等因素相关。而限制因素是指资源消耗模型、用户需求等。划分目标可以是最低消耗,最高性能,最健壮或者最安全等。所以弹性划分方法应该基于以上因素进行划分。根据弹性应用程序模型,本文提出了基于应用程序的划分消耗图、弹性架构、改进的(K+1)粗粒度划分算法以及划分消耗模型。改进的(K+1)粗粒度划分算法用于划分消耗图。图中点的权值是由移动设备的消耗和迁移到云服务器的消耗之和所组成。这两部分消耗可以由一些上下文信息来表示,其中包括执行时间、当前电量、网络环境以及划分概率。划分概率是由历史数据统计分析得出的。这些信息的权重可以根据上下文感知的层次进行调节。为了适应动态环境的变化,该算法可以进行再次划分。划分消耗模型被用于存储和计算上下文信息。实验将OCR项目部署在该架构下。在不同的网络环境和输入条件下,该算法都得到了很好的执行效果。
[Abstract]:In recent years, mobile applications and mobile devices have made great progress. But mobile devices are still unable to run data-intensive applications compared with PC. These devices are limited by battery power, screen size, wireless communication, etc. In order to overcome these limitations, we can use mobile cloud computing technology. Mobile cloud computing can be defined as an extension of cloud computing. Cloud computing has powerful computing power, massive data storage and provides a new business model, that is, computing resources to provide services to users. Mobile cloud computing hardware resources also include mobile devices. A distributed infrastructure can be built using the collective sensing, network connectivity, storage, and computing capabilities of different mobile devices. Migrating mobile applications to mobile cloud platforms will face a problem of elastic partitioning. Elastic partition is a complex problem. The results of elastic partitioning are affected by many factors, including input factors, limiting factors and dividing objectives. Input factors are related to battery power, network status, device load, cloud load, and current execution performance. The limiting factors are the resource consumption model, user demand and so on. The goal can be the lowest cost, the highest performance, the most robust or the safest, etc. So the elastic partition method should be based on the above factors. Based on the elastic application model, this paper presents an application-based partition consumption graph, an elastic architecture, an improved (K1) coarse-grained partitioning algorithm and a partitioning consumption model. The improved (K 1) coarse-grained partitioning algorithm is used to partition the consumption graph. The weights of the midpoint are composed of the consumption of mobile devices and the consumption of migration to cloud servers. These two components of consumption can be represented by some context information, including execution time, current power, network environment, and partition probability. The probability of division is derived from the statistical analysis of historical data. The weight of this information can be adjusted according to the context-aware hierarchy. In order to adapt to the change of dynamic environment, the algorithm can be divided again. The partition consumption model is used to store and compute context information. The experiment deploys the OCR project under this architecture. Under different network environment and input conditions, the algorithm has a good performance.
【学位授予单位】：云南大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TP3

【相似文献】