Design And Implementation Of Cloud Platform Scheduling System Based On LVS

With the expansion of the cloud computing platform,its internal tasks become increasingly rich,resulting in the growth of the number of service nodes in the resource pool.Reasonable tasks and resource scheduling can speed up the implementation of the task,thus the user experience can be improved.Therefore,the scheduling system has been the focus of industry word.The DR model in the LVS scheduling system has a very good anti-concurrent capability,so it has been widely used.But the algorithm which used in the DR Model use static parameters such as number of the connections as the scheduling basis,which do not take into account the real-time status of the node in the resource pool.And because the LVS is used in the fourth layer forwarding technology,it can not get the task characteristics.In the complex environment of cloud computing,it is not good for the large task scheduling.Finally,in LVS and other traditional scheduling systems,the detection of the health status of the nodes in the resource pool is relatively simple,but to determine whether normal.They only use the normal nodes,and the measures for the fault node is to separate it from the resource pool until the administrator to restore the node.That leads to a waste of resources.Based on the LVS scheduling system,DFSS is designed and implemented in the laboratory cloud platform environment.First of all,the DFSS make full use of the newly proposed Google third generation scheduling system Omega by adding a resource sharing module for itself,which make modules in the cluster can get overall situation of the cluster.Then,the dynamic feedback algorithm is designed and implemented by using the shared resource module,so that the LVS can take full account of the real-time status of the nodes in the resource pool when the task is allocated.Then,according to the characteristics of complex tasks in the cloud environment,we set up a special module to classify the complex tasks,so as to achieve a reasonable scheduling.Then,We improve the health detection mechanism of traditional scheduling system.Based on the idea of autonomous computing,the self-detection and recovery function of the nodes in the resource pool is realized by the software log,which improves the autonomy of the system.Finally,based on the shared resource module,we has designed and implemented the control center module,which is responsible for monitoring the overall state of the cluster.When large number of nodes in the cluster are in extreme idle state,the control center will take the strategy to close some idle machines to reduce the overall energy consumption of the cluster.Moreover,if the control center detects that all nodes in the resource pool are in a state of extreme tension for a long time,the relevant strategy will be adopted to apply the virtual module to add resources and join the resource pool.