Novel Energy Aware Resource Allocation and SDN Integrated for Reducing Power Consumption in CLOUD

Abstract

newlineIn recent cloud services, resource allocation and provisioning is significant for assigning the available resources in energy efficient way. Resource management becomes challenging due to high energy consumption at data center, Virtual Machine (VM) migration, high operational cost and overhead at data centers. Many researchers have concentrated on research management using virtualized technologies with several algorithms such as optimal-Multi-objective particle swarm optimization, Dynamic Power Saving Resource Allocation (DPRA), Least Square Regression, etc. However, these previous processes are difficult for allocating resources accurately to cloud users in order to meet their requests and also difficult to provide QoS. Energy efficient resource allocation is essential to achieve user satisfaction and maximize the profit for cloud service provides. In this research, we concentrate on resource management problems and address those problems by involving various novel processes with heuristics, authentication and virtualization techniques. Dynamic scheduling with load balancing is a very important task in cloud computing that distributes the load across physical nodes based on current load information. Markov chain model assisted with Particle Swarm Optimization and ETC matrix with PSO dynamically allocates the VM resources to appropriate input requests by exploiting knowledge of future prediction value of host load. newlineIn order to provide high QoS for cloud applications, SLA based resource optimization techniques are considered with SLA objectives such as deadline, cost, storage and bandwidth. SLA based priority clustering algorithm is designed for satisfying SLA constraints which improves the resource optimization, energy consumption and reduce the SLA violations. Further, Queuing is an important mechanism which allows the scheduled users in sequential way to access the resources. We design the M/M/c/K queuing model which supports several users with individual server to avoid overhead in data center. Resource information of VM is collected from PMs that are characterized by power models such as CPU, I/O and memory usage. This information collection process enhances the utilization of resources and reduces the power consumption of each and every resource. Based on resource utilization, resource allocation and migration process are performed. VM prediction is essential for reducing the number of active physical servers and improving the energy efficiency in cloud data centers, with the goal to reduce the total power consumption. VM migration is processed with respect to their statuses newlinesuch as idle, running, busy, power off and sleep which are considered to reduce the power consumption while performing the userand#8223;s task. newlineWe integrate the cloud with SDN which reduces the conventional network problems in resource allocation and provides the centralized control for accessing the resources to cloud users. SDN enabled cloud is introduced for enhancing the features such as high throughput, auto addressing, network virtualization, fault detection, traffic classification, load balancing and routing. Here, user authentication is also considered for allowing authorized users to access the cloud resources. Therefore, Multi Factor Authentication (MFA) is involved to verify the authentication level of users. SDN controller is an open flow centralized controller which manages the resource optimization process through establishing connection between switches. In SDN controller, queuing mechanism, resource status information, VM prediction, VM allocation and VM migration are performed. Finally, our experimental results provide the better performance in throughput, execution time, cost, power consumption, resource utilization, energy consumption, SLA violation rate and number of VM migrations with comparing previous resource allocation process. Our overall process improves the resource utilization of computing resources, reduce energy consumption under workload independent SLA constraints and reduce the resource management problems in energy efficient way.