Performance Analysis And Scheduling Stochastic Requests To Heterogenous Cloud Resources

The arrival patterns of requests for consumers follow different distributions such as Poisson distribution,general distribution with different constraints such as different patience and deadlines.The service resources are usually heterogeneous servers that follow exponential distribution for single queue scenarios or multiple queue scenarios.Different complex queueing models are constructed in terms of different arrival patterns and constraint combinations.How to determine the number of servers and scheduling requests reasonably to optimize response time and power consumption,rental cost,system energy consumption,and profit for service providers is one of the critical problems in cloud computing.According to different patience and deadlines of stochastic requests,it is essential to study performance analysis and scheduling optimization methods for heterogenous servers in single queue models and multiple queue models.The novelties of this dissertation is described as follows.· Performance analysis and optimization for cloud service stochastic requests with a single queue is considered.According to heterogeneous servers and stochastic cloud service requests,a single queue model is constructed.A state space of system states is determined in terms of the setup time,states,and types of heterogeneous servers.The Markov process calculates the rejection probability of the cloud system.According to the system availability,the number of servers is determined by a binary search method and the initial types of servers are obtained.The policy evaluation algorithm is proposed to evaluate the performance of the selected server configurations.An iterative improvement method is proposed to determine the best servers to select for the considered objective.The performance of the system parameters on algorithms' performance using the analysis of variance is evaluated over alarge number of random instances.The proposed algorithm is compared with other related algorithms over a large number of random and real instances.The results indicate that the proposed algorithm is much more effective than the other algorithms within acceptable CPU times.· Performance analysis and optimization for cloud service stochastic requests in a single queue with different impatience are considered.For impatient consumers with different patience,a single queue model is constructed to minimize the rental cost.According to the configuration of heterogeneous servers and queue capacity,the system state space is constructed.The actual waiting time of requests is calculated according to the Markov process.The rejection probability is calculated in terms of maximizing waiting time and impatient consumers.By comparing the system's rejection probability to a given system availability,it can be checked whether selected servers satisfy the constraints.A cost minimization algorithm framework is proposed to select suitable servers that satisfy impatient consumers.The performance of the system parameters on algorithms' performance using the analysis of variance is evaluated over a large number of random instances.The proposed algorithm is compared with other related algorithms over a large number of random instances.The results indicate that the proposed algorithm is much more effective than the other algorithms within acceptable CPU times.· Performance analysis and optimization for cloud service stochastic requests in a single queue with deadline constraints are considered.According to the deadline constraints for stochastic requests and elastic queue capacity,a single queue model is constructed with an elastic queue capacity.The queue capacity and state space are determined based on the heterogeneous servers and deadline constraints.According to the Markov process,it proves that the rejection probability changes because the queue capacity changes.A strategy is proposed to optimize the server configuration and queue capacity in terms of heterogeneous servers and deadlines.The energy minimization algorithm is proposed for cloud service stochastic requests with a single elastic queue.The performance of the system parameters on algorithms' performance using the analysis of variance is evaluated over a large number of random instances.The proposed algorithm is compared with other related algorithms over a large number of random and real instances.The results indicate that the proposed algorithm is much more effective than the other algorithms within acceptable CPU times.· Performance analysis and optimization for cloud service stochastic requests in multiple queues with patience are considered.According to distributed heterogeneous servers and stochastic consumers with patience,a multiple queue model is constructed.A request stream splitting algorithm is designed to reasonably split the arriving requests into multiple queues to minimize the response time.According to the server property in each queue,an allocation algorithm is developed to optimize the requests' response time further.For heterogeneous servers,the number of servers is selected according to the variable neighborhood search to maximize total profit for service providers.The performance of the system parameters on algorithms' performance using the analysis of variance is evaluated over a large number of random instances.The proposed algorithm is compared with other related algorithms over a large number of random and real instances.The results indicate that the proposed algorithm is much more effective than the other algorithms within acceptable CPU times.