面向交通感知数据处理业务的虚拟机调度系统

海量、实时、持续、多样的交通感知数据对承栽其处理业务的IT设施提出了更高的要求,但现有云基础设施服务中采用的虚拟机调度方案在承栽交通感知数据多处理业务时，会出现负栽倾斜、负栽不均及资源利用率不高的问题。为此,提出一种对交通感知数据处理业务敏感的虚拟机调度策略并设计了相应的调度算法，给出了系统的设计与实现,该系统可以适应计算密集和数据密集不同特征交通处理业务的承栽，同时能够实现能耗约束下的负栽平衡调度。通过基于实际交通应用场景的实验表明，该系统可在保障所承栽业务的前提下实现虚拟机的负栽均衡和优化节能。

Virtual machine scheduling system for multiple traffic sensor data processing applications hosting

Massive, real-time, continuous and various traffic sensor data put forward higher requirements on the underlying IT infrastructure. While when directly applying the virtual machine scheduling strategy of current Cloud IaaS (Infrastructure as a Service) services to host different processing applications on traffic sensor data, the phenomenon of unbalanced load among physical machines often appears. This paper puts forward a hosting application-sensitive scheduling strategy on virtual machine and corresponding greedy algorithm, aiming at solving the problem of dynamic virtual machine scheduling according to the changing needs of sensor data processing. The virtual machine scheduling system based on above strategy and algorithm is also designed and implemented. Real application cases have shown that the virtual machine scheduling system proposed in this paper can be used to realize the goal of load balancing and optimized energy saving.