论文部分内容阅读
目的:为提供持续性的按需服务,移动云计算系统必须确保在任何时间和任何地点的可用性。然而当系统规模巨大、关联关系复杂时,如何实现移动云计算系统可用性的快速分析,是本领域一项富有挑战性的工作。本文目的是利用最近提出的流近似(fluid-flow approximation)技术来实现一种能应用于移动云计算系统部署之前的、快速的服务可用性分析方法。创新点:由于移动云计算系统负载水平不同、配置部署不同和随机干扰因素,基于实测的方法很难具有代表性;基于随机模拟的方法会随着模拟规模增大和精度提升而计算时间剧增;基于状态空间的方法在系统规模巨大时将面临严重的状态空间爆炸问题。本文方法通过将状态空间转化为常微分方程组求解,可以避免状态空间爆炸,实现移动云计算系统可用性的快速分析。方法:定义了包括服务反应时间(response time of service)、节点最小感知时间(minimum sensing time of devices)、最少选取节点数量(minimum number of nodes chosen)、动作吞吐量(action throughput)等四个关键指标。通过上述指标来分析移动云计算系统服务可用性的变化,并对系统初始条件、模型核心参数的影响进行讨论。结论:本文提出的服务可用性分析方法能够适用于移动云计算系统完全部署之前,可以用于系统设计阶段的改进。并且与基于随机模拟方法和状态空间方法相比,时耗更低。
Purpose: To provide continuous on-demand services, mobile cloud computing systems must ensure availability at any time and anywhere. However, when the system is huge and the correlation is complex, how to quickly analyze the availability of mobile cloud computing system is a challenging task in this field. The purpose of this paper is to use the recently proposed fluid-flow approximation technique to achieve a rapid service availability analysis method that can be applied to mobile cloud computing system deployment. Innovation: Due to different load levels of mobile cloud computing systems, different deployment configurations and random interference factors, it is very difficult to have representativeness based on measured methods. The method based on stochastic simulation will increase dramatically with the increase of simulation scale and accuracy. The state space-based approach faces a serious state-space explosion problem when the system is large-scale. In this paper, the state space can be transformed into an ordinary differential equation solution to avoid the explosion of the state space and achieve the rapid analysis of the availability of the mobile cloud computing system. Methods: Four keys including response time of service, minimum sensing time of devices, minimum number of nodes chosen, and action throughput are defined index. Through the above indexes, we analyze the change of service availability of mobile cloud computing system, and discuss the influence of system initial conditions and model core parameters. Conclusion: The service availability analysis method proposed in this paper can be applied to improve the system design stage before the mobile cloud computing system is fully deployed. And the time consumption is lower than that based on stochastic simulation method and state space method.