Research On Cloud Host Monitoring Service In Cloud Migration Of Railway Business System

With the application of cloud computing technology in the construction of railway informatization,it has become a trend to migrate railway-related services to the cloud.However,this poses a greater challenge to the private cloud platform for railways.In order for the railway business to provide normal external services on the cloud,it is first necessary to ensure the stable operation of the underlying infrastructure in the platform,the cloud host.At present,due to the large number of host devices in the cloud platform,and difficult maintenance and management,when the host is abnormal,the problem cannot be located and handled in a timely manner.Moreover,due to the unreasonable resource allocation,the load imbalance between the hosts and the waste of resources on the hosts are caused.In view of the above situation,this article has carried out research on the host monitoring and resource scheduling in the railway private cloud platform.The specific contents are as follows:In terms of host monitoring,this paper proposes a distributed architecture design of the monitoring system based on Zabbix open source software,which can achieve horizontal expansion of the system.For the "cloudy" environment of the railway,an automatic deployment method of the monitoring system is proposed based on Docker and Ansible technologies.This method can automatically deploy monitoring services on multiple hosts.At the same time,through the secondary development of Zabbix,increased monitoring of container services on the host.Experimental results show that the use of automated deployment methods improves the deployment and start-up efficiency of the monitoring system.And the monitoring system after startup can obtain the status information and resource usage information of the container service on the host.In terms of resource scheduling,this paper proposes a particle swarm optimization algorithm with a crossover operator for the optimization of virtual machine allocation.Through this algorithm,it finds the best mapping scheme from virtual machine to physical machine.It also models the optimization goal of the allocation problem,and proposes a resource waste model and a load balancing model,which are used as the main basis for measuring the merits of the virtual machine placement scheme.Experimental results show that the solution found by the newly proposed algorithm is closer to the optimal solution of the allocation problem,and it is also conducive to resource utilization on the host and load balancing between the hosts.Finally,this paper proposes a dynamic migration mechanism for virtual machines in combination with the monitoring system and distribution algorithm implemented above.The mechanism realizes the detection of overloaded and underloaded hosts by monitoring services and multiple thresholds.In response to the selection of virtual machines to be migrated from overloaded hosts,three migration mechanisms are proposed,namely: minimum utilization strategy,maximum utilization strategy and CPU memory ratio strategy.And through the new migration energy consumption model,the choice of migration mechanism is realized.Regarding the allocation of out-of-place virtual machines,a newly proposed algorithm is used.Experimental results show that this dynamic mechanism can balance the load between hosts and reduce the energy consumption for migration.