Research And Implementation Of Container Scheduling On Container Cloud Platform

Container is a new virtualization technology.Deploying applications in containers is more resource efficient,and easier to migrate than virtual machines.The traditional virtual machine scheduling strategy is not suitable for container technology because of the differences between container and virtual machine in the underlying and resource granularity division.For the container itself,considering its running environment,the container can be either deployed to the hosts or the virtual machines,so the scheduling problem of the container becomes more complex.In the cloud environment,the load changes frequently.How to predict the load value and allocate resources according to the predicted results is an urgent problem to be solved.Based on the above analysis,this paper analyzed the current research situation of container scheduling problems,summarizes the existing problems in the current research,designs and implements several container scheduling optimization strategies.(1)Most current container scheduling studies only consider a single objective.Based on this,a multi-objective optimization strategy is proposed based on resource waste caused by node resource fragmentation,energy consumption of cloud data center,resource utilization.The corresponding evaluation function is given for each factor,the complementarity between the container node and the physical machine node is calculated by Normalized Euclidean distance,the weighted sum method is used to transform the multi-objective problem into a single objective,and finally the optimal solution is obtained by integer programming.The experimental results show: compared with the classical algorithms,the algorithm proposed in this chapter has the least comprehensive cost and can effectively reduce energy consumption and improve resource utilization.(2)the container deployment problem is combined with the packing problem,and the containers are deployed to the virtual machine by using the container-virtual machine-physical machine two-layer scheduling architecture for modeling.Due to the ant colony optimization algorithm is slowly convergence and easy falling in local best,The state transition probability and pheromone update rule are improved,and combine the simulated annealing algorithm to solve the problem.The experiments result show that this strategy can reduce the energy consumption of data center to some extent.(3)during the container migration,if only considering the current load state,the container with a high current load but a low load at the next moment may be selected for migration.It doesn't work in this case.Therefore,this paper improves BP neural network algorithm through particle swarm optimization to predict container load.For container migration,container selection takes into account both the current load state and the predicted load.Experiments show that the improved BP neural network based on particle swarm optimization has better prediction effect than the traditional BP neural network.Meanwhile,the container selection strategy based on load prediction proposed in this paper can effectively reduce SLA conflicts.