| In cloud computing environment,Infrastructure as a Service(Iaa S)recently emerged as a new archetype for hosting and delivering computing resources including computing capacity,storage capacity and network capacity to cloud users over the world on the internet demand.However,the increasing number of cloud users leads to the exponentially growing number of user applications,which requires very large amount of computing resources in cloud data centres to respond to the service level agreements(SLA)between cloud users and cloud providers.In contrast,large cloud data centres can lead to an increase energy consumption and several migrations,which negatively impact on performance of resource management in Iaa S Cloud.Thus,the contrast between the growing number of user applications and the growth of computing resources,becomes a great challenge of resource management.To overcome this challenge,this thesis analyses how to deal with the issues caused by the challenge,including: massive and complex computations,data massive,various computational environments,and the increase in energy consumption and migration while maintaining low computation time and high level of SLA.Hence,this thesis investigates on low-time complexity computation methods based on the integration of load balancing and scheduling schemes,which deal with each issue to make the methods very useful in realistic computational infrastructures.The main objective is to propose a quick,efficient and simple global scheduler which combines the techniques of load balancing and scheduling schemes to optimize resource allocation and resource provisioning in Iaa S Cloud resource management.(1)Cloud computing faces with the increasing number of compute-intensive applications under the limited computing power and number of heterogeneous resources,where cloud computing breaks down applications into large tasks and small tasks.That leads to high computation time,and then to the inefficient resource allocation due to the high task delay time and lack of sufficient computing resources.Hence,this thesis proposes a low-time complexity heuristic approach for integrated approach of load balancing and task scheduling to handle the enormous amount of computation demands subject to low computation time and limited capacity of heterogeneous resources.The proposed approach focuses on finding an efficient method,which calculates a common and minimum completion time among allocated heterogeneous virtual machines to minimize the completion time of each virtual machine,and the overall computation time in a heterogeneous environment.(2)In cloud resource management,some tasks require a large number of network bandwidths to be executed due to their massive data.However,network bandwidth is limited to satisfy the increasing number of data-intensive tasks.That can lead to high computation time,high execution cost and fewer profits due to high network load,and then to the inefficient resource allocation.Hence,this thesis proposes a fast and efficient scheduling heuristic approach based zero imbalance mechanism on handling the massive data of tasks subject to low completion time and limited capacity of heterogeneous resources,including limited network bandwidths.The proposed approach implements a zero imbalance mechanism,which achieves a high level of load balancing to meet requirements both cloud users and cloud providers.Besides,the highest level of load balancing can improve the availability and scalability of computing resources.(3)In cloud computing,computation time and performance depends on the number and capacity of computing resources,which defines a particular type of computational environment to handle the increasing number of compute-intensive tasks and data-intensive tasks.However,it is difficult to define an appropriate computational environment due to the unpredictable of number,type and size of user tasks.Hence,this thesis proposes a novel binary particle swarm optimization(BPSO)with low-time complexity and low-cost,which performs a combination of task scheduling and load balancing adapted on various computation environments,because of its randomness for optimal resource allocation.Also,during the initialization stage,the proposed BPSO method does not only use a random method but also makes use of the proposed heuristic algorithm to initialize a particle among a population of particles to increase the performance of resource management.To achieve the proposed approach,this thesis focuses on finding a method which calculates a reference for each particle to facilitate the convergence toward the optimal solution and to accelerate the search exploration in binary space so that it results in low time complexity.Besides,the thesis also focuses on improving the updating method for particle position concerning load balancing strategy.(4)Due to the advances in electronics and growth finite heterogeneous resources in cloud data centres,the computing resources not only produce high energy consumption,but they can also activate a great amount of migrations of virtual machines which can lead to high network load and high cost of SLA violation.Although,high energy consumption can help to meet the growth of user demands according to SLA,it also negatively impacts the human environment and the performance of resource provisioning.Hence,this thesis proposes low-time complexity dynamic VM consolidation approach based load balancing using Pearson correlation coefficient,which optimally schedules and balances computing resources,to minimize the tradeoff between energy consumption,the cost of SLA violations and the number of migrations in heterogeneous cloud data centers.Unlike to previous researches,the proposed approach does not only focus on CPU,but it also considers RAM memory and bandwidth which can affect the efficiency of resource management. |
