| With the development of cloud computing and Internet of things technology,smart parking cloud based on SaaS mode has been widely used,effectively breaking the current situation of information asymmetry,high cost of manual management and utilization rate of parking lot in traditional parking industry.Multi tenancy is the core mode of parking cloud.Since parking lot owners share a set of customized smart parking services deployed on the cloud,it significantly reduces the economic cost of providing smart parking services of the cloud service provider.However,the sharing of cloud computing resources leads to service performance competition among tenants.Therefore,how to ensure that all tenants in the cloud have high-quality parking services becomes the most concerned issue for cloud service provider.Unlike traditional SaaS applications,smart parking cloud integrates IoT devices through cloud computing to provide intelligent parking services for end users,covering two application scenarios:Web and IoT.Firstly,according to the common characteristics of the multi-tenant model of smart parking cloud,combined with container technology and micro-service architecture,this paper proposed a micro service container deployment scheme for tenant SLA.Then,according to the features of smart parking cloud in web and IoT scenarios,performance isolation algorithms for microservice cluster in Web scenario and subscribe/publish mode in IoT scenario are proposed respectively.Specifically,the research work of this paper is as follows.1)A micro-service containerized deployment scheme for tenants’ SLA is designed.According to the business scale of parking lot,the corresponding level SLA agreement is signed with the tenants,and the container technology is used to meet the performance requirements of tenants in different level.This scheme achieved the performance isolation of tenants between different SLA levels.2)The performance isolation algorithm is implemented for the micro-service cluster in the web scenario.Based on the reverse proxy technology,this algorithm schedules HTTP requests from different tenants through the statistics of tenant resource consumption,which provides effective performance isolation of web services.Compared with the fair queue algorithm,this algorithm can better balance the resource occupation of the tenant when the resource consumptions of the tenants’ requests are different.3)The performance isolation algorithm is implemented for the subscribe/publish mode in the IoT scenario.Aiming at the MQTT protocol which is commonly used in the field of IoT,a multi-tenant data flow scheduling algorithm is proposed.When the IoT devices communicate with the cloud platform,the algorithm can effectively reduce the performance interference between different tenants’ devices,and achieve effective performance isolation for the tenants’ IOT devices.Compared with token bucket algorithm,this algorithm achieves stronger performance isolation.4)According to the scheme and algorithms proposed in this study,the smart parking cloud performance isolation system is implemented,and experiment of multi-tenant high concurrency scenario is carried out for the smart parking cloud platform after the deployment of the system.By increasing the concurrency of HTTP requests and the IoT massages from some tenants,the working load of cloud services caused by the tidal effect of the parking lot is simulated.The experiment results show that the performance isolation algorithms proposed in this study can effectively reduce the performance competition between tenants in the two scenarios of web and IoT,and achieve performance isolation between tenants. |
