MRGP-based Availability Analysis Of Application Service In A Virtualized System

System virtualization enables multiple virtual machines simultaneously running on a single physical machine.As an important technology for improving IT efficiency,it is widely used in academia and industry.However,with the long-term operation,the virtualized system will inevitably suffer from software aging,resulting in decreased performance or even downtime,and the application service running on it will eventually violate the service level agreement(SLA),which causes huge economic losses.Therefore,in order to meet the user's demands for application service quality,how to improve the availability of application service in the virtualized system has become a common concern of service providers and users.However,in the existing studies on improving availability,either only adopting two-level rejuvenation in the virtualized system and not considering the application service level,or even if the three-level rejuvenation is considered,it is only modeled by continuous-time Markov chain,and such models have certain limitations in practical research.Based on the above problems,this paper studies the aging problem of a virtualized system with three levels(AS,VM and VMM),and applies the Markov regeneration process theory to characterize and analyze it.Measurement-based rejuvenation as a kind of software rejuvenation can reduce the impact of software aging on application service availability in the virtualized system,but its effectiveness depends on the rejuvenation strategy,which defines when and where to rejuvenate,and what kind of rejuvenation technology is triggered.This paper mainly studies the impact of two rejuvenation strategies on application service availability in a two-machine virtualized system,and analyzes the approximate optimal inspection time interval corresponding to the maximum availability in each strategy.The paper first applies the Markov regeneration process to build an analysis model of the three-level rejuvenation strategy for the virtualized system with two physical machines,that is,the rejuvenation technology is deployed at each level of AS,VM,and VMM.Specifically,rejuvenation technologies including AS restart,VM restart,VMM restart,VM failover,and Live VM migration are used to deal with aging.Then,in order to further verify the effectiveness of the three-level rejuvenation strategy,a model is built for the system under a two-level rejuvenation strategy,where the AS level does not take active rejuvenation actions,and only the two levels of VM and VMM are rejuvenated.Through the experimental analysis of model,the quantitative relationship between AS availability and system parameters is studied,and the approximate optimal inspection interval is determined.The experimental results show that the change trend of the availability with the inspection interval increases first and then decreases.The optimal inspection interval is the value that maximizes the availability.Moreover,compared with the two-level rejuvenation strategy,three-level software rejuvenation strategy can produce higher AS availability and lower loss probability,with unavailability and loss unavailability reduced by 63.04%and 47.54%,respectively.These analysis results have very important guiding significance for service providers to design and provide specific SLA systems and evaluate existing functions.