Research On Adaptive Monitoring And Availability Enhancement For Mission-Critical Computers

The mission-critical computer system means a computer system carries out mission-critical tasks,including Internet banking system,operation and control system for railway and aircraft,electrical system and so on.These systems have a common problem that it might cause a huge damage to enterprises and even the society when these systems are down.Existing monitoring systems and availability enhancement methods can not fully play the maximum performance of the key tasks of the computer.Therefore,it's very significant to find a way to maximize the availability of the mission-critical system with limited hardware resources,which keep the mission-critical system in high availability.To optimize the availability of the mission-critical computer system,a new method of adaptive monitoring and the availability enhancement technology is proposed.The computer system can maintain very high availability in the low load condition,but with the increasing of the workload,the availability of mission-critical system will be still declining.In this paper,an adaptive monitoring method for mission-critical computer is proposed,which will effectively reduce the impact of monitoring function in high-load condition by changing the monitoring scheme according to the relative availability.The relative availability could be calculated by monitoring the load of each virtual machine.According to the monitoring data,three monitoring schemes are adopted to ensure system availability.Most of the existing methods of resource dynamic allocation are based on threshold value,which brings problems that resource allocation is unbalanced and not timely.In this paper,CPU shifting allocation method,CPU rotation allocation method and memory shifting allocation method are proposed by combining thresholds with the change rates of computing resources.These methods improve availability of mission-critical system by realizing the dynamic allocation of resources quickly and effectively,which solve the problem that the dynamic allocation of resources is not timely and improve the availability of mission critical computers.Workload is an important indicator to judge the availability of mission-critical computer system,the availability of the system will decrease significantly with the increase of workload.From the view of workload,a method to predict the critical tasks based on SVM is proposed.Firstly,according to CPU utilization,memory utilization,network load rate and I/O load rate,a predicting model of availability based on SVM is proposed.Secondly,this paper predicts availability of the mission-critical computer by using historical monitoring data,after completing parameter optimization of SVM kernel function.By predicting the availability of mission-critical system,better management and maintenance of the system will be taken.