DANCE:一种面向云-端动态集成的服务适配方法

边缘计算可以通过将计算移到边缘设备上来提高大型物联网流数据处理质量以及降低网络运行成本.对于流数据处理,边缘设备通常只有有限的计算能力和存储能力,显然不能支持所有的实时流数据查询和处理.本文尝试引入服务并在边缘和云之间灵活地划分服务来实现云-端集成,云服务和端服务之间通过事件机制进行服务适配.物联网动态环境中,云-端服务的动态适配是使云基础设施和端设备间无缝集成的关键.动态集成背景下的服务适配需要把握适配时机来应对端服务适配请求的不确定性和非完全适配等难题.针对这一问题,论文提出了一种面向云-端动态集成的服务适配方法(Dynamic Adaption cloud Services with Edge Services,DANCE).这种方法的主要贡献在于:将云服务实例和端服务实例之间的适配问题建模为二分图顶点之间的动态匹配问题,同时结合排队论中的M/M/c/∞模型对二分图最优匹配Kuhn-Munkres算法进行了优化改进,保障适配过程中端服务实例的全局平均请求响应时间最小.最后,基于真实的电能质量监控案例和数据,验证了本文方法的有效性.

DANCE:A Service Adaption Approach for the Dynamic Integration of Cloud and Edge

In IoT,large-scale sensor data are continually generated in the form of data stream.These data are usually infinite,noisy,of low-value density,and orderless on a broad panorama.To provide high-quality services,comprehensive analyses of big stream data based on cloud are essential.Although having powerful potential,cloud computing has faced increasing computational challenges like performance,cost,energy consumption,and service qualities because of the exponential growth of big IoT data.Edge computing may improve the processing quality of big IoT stream data and reduce network operational costs by moving computation onto the edge.While the edge equipment usually has very limited computing power as well as storage ability,and apparently cannot support all the processing of big and real-time stream data.Our previous work has introduced services and provided a flexible division of such services between the cloud and the edge for the cloud-edge integration.The service adaptation between the cloud service and the edge service can be accomplished through the event mechanism.It is the key to enabling the seamless integration of cloud infrastructure and edge equipment.However,in IoT context,the adaption between cloud services and edge services faces challenges because of the uncertainty of edge service request arrivals as well as incomplete matching.Firstly,compared with the cloud infrastructure,edge equipment can join or quit at any time due to their own flexibility.Secondly,the capability of the edge equipment will constantly change with using which can result in dynamic change of the state of the edge services.Finally,driven by the event,the moment of adaptation request for edge service is uncertain because of the peculiarity of the large-scale streaming data.It is because that the generation and arrival of events on edge equipment are uncertain.Thus,the dynamic service adaption in the context of dynamic integration needs to grasp the right moment of adaption to face the challenges of uncertainty of edge service request arrivals as well as incomplete matching.To solve this problem,this paper puts forward a service adaption approach for the dynamic integration of cloud and edge,called as DANCE.The main contributions include:we transform the dynamic service adaptation problem into the improved maximal weight matching model with a dynamic bipartite graph.At the same time,we modify the M/M/c/∞ model in the queuing theory.The modified M/M/c/∞ model is used to optimize the Kuhn-Munkres algorithm for bipartite graph to minimize the average response time for the request of edge service.DANCE guarantees the minimum global average request response time for the edge service instances during the adaptation process.Finally,based on a real dataset from the case of China’s State Power Grid,the effectiveness of the proposed approach is demonstrated by examining a series of simulation experiments.Experimental results show that,DACNE can perform better than us previous work that does not consider dynamic adaption,and the total average response time of DANCE is 24.3%less than us previous work.Experimental results verified the effectiveness and efficiency of our approach.