| Virtualization technology promotes the rapid development of cloud computing technology and accelerates the arrival of the era of Internet of Everything.In a system-virtualization-based system,virtual machine monitor plays an important role.It provides services for the upper virtual machine by virtualization of hardware.However,virtual machine monitors are software that is subject to aging after a long-running continuous process,which will lead to software performance degradation,thus increasing job completion time(the total time required to complete a job)and decreasing application service availability(percentage of normal service provision time for the system).There are various software rejuvenation techniques,including restart techniques of system and service,failover technique,migration technique and so on.Software rejuvenation policies determine how these techniques are scheduled and how they work together.Therefore,it is of significance to study the impact of software rejuvenation policies on job completion time and application service availability.Researchers have achieved fruitful results on the design and effectiveness analysis of software rejuvenation policies.However,these researches have the following problems.Firstly,the designed model cannot quantitatively analyze job completion time and application service availability at the same time for software rejuvenation policies.Secondly,no research has been done on virtualization systems with migration technique deployed.Thirdly,the system which time distributions are general distributions cannot be quantitatively analyzed.This thesis aims to construct two software analytical models in order to analyze active return-migration software rejuvenation policy and passive return-migration software rejuvenation policy for the virtualization system with restart technique and migration technique deployed,under the condition that the aging time,failure time and rejuvenation time of physical server and migration time of virtual machine follow general distribution.Analytical experiment is used to evaluate application service availability and job completion time and to obtain the optimal migration trigger time corresponding to the above indicators.The main contributions are summarized as follows:(1)Constructing the active return-migration software rejuvenation policy analytical model and analyzing quantitatively its evaluation indexesFirstly,the main physical server's behaviors after software aging are analyzed in detail for system with active return-migration software rejuvenation policy.Based on active return-migration software rejuvenation policy,this thesis deploys migration technique and restart technique for system.Then,the analytical model under the corresponding active return-migration software rejuvenation policy is constructed based on the semi-Markov process.Formulas of job completion time and application service availability are deduced for quantitative analysis.Finally,numerical analytical experiment is used to evaluate the impact of system parameters on the application service availability and job completion time.The experiment also obtains the maximum application service availability and the shortest job completion time and their corresponding optimal migration trigger time.(2)Constructing the passive return-migration software rejuvenation policy analytical model and analyzing quantitatively its evaluation indexesFirstly,main and backup physical server's behaviors after software aging are analyzed in detail for system with passive return-migration software rejuvenation policy.Based on passive return-migration software rejuvenation policy,this thesis deploys software rejuvenation techniques for system.Secondly,the corresponding software rejuvenation policy analytical model is constructed.This thesis quantitatively analyzes job completion time and application service availability and obtains optimal migration trigger time.Finally,two policies are compared from these two aspects. |
PDF Full Text Request