Techniques to Minimize Energy Consumption in Cloud System

Abstract

Cloud computing is being widely applied to a variety of large size computational problems.
These computational environments consist of many heterogeneous computing modules;
these modules incorporate with each other to implement the solution of various problems.
A typical cloud deployment consumes a significant amount of energy, and higher energy
consumption has an adverse impact on the environment. Reducing energy consumption in the cloud environment is both a research and an operational challenge for the current research community and industry. The objective of this research is to minimize the energy consumed by the cloud system, in particular, considering the execution of tasks (service requests) with the help of virtual machines. A survey of the state-of-the-art in an energy-efficient cloud computing system is presented. In this thesis, we have used four different approaches: (i) task allocation, (ii) virtual machine consolidation, (iii) virtual machine selection using Dynamic Voltage and Frequency Scaling (DVFS), and (iv) resource allocation in mobile cloud system to optimize energy consumption of the cloud system. All the proposedalgorithms are simulated with the help of CloudSim simulator.

The task allocation problem is a well known NP-complete problem. We have presented
two different approaches of task allocation to optimize the energy consumption and
makespan of the cloud system. The first approach deals with three task allocation algorithms based on metaheuristics techniques namely, Particle Swarm Optimization (PSO), Binary PSO (BPSO), and BAT. In the second approach, a deterministic Adaptive Task Allocation Algorithm (ATAA) is proposed to allocate tasks to the cloud system. The task allocated to the cloud system is represented with the help of an ETC (Expected Time to Compute) matrix. The ETC matrix holds the time required to compute a specific task on different Virtual Machines (VMs). The simulation is carried out to compare the performance of three proposed metaheuristic based task allocation algorithms by varying the number of VMs and tasks. And, also the performance of the proposed Adaptive Task Allocation Algorithm (ATAA) is analyzed by comparing with the random, and Round-Robin (inbuilt algorithm in CloudSim) algorithms. Simulation results indicate in favor of the proposed scheme (ATAA).

In a cloud system, VM consolidation deals with the allocation of VMs to hosts. In this thesis, a task-based VM-consolidation algorithm is proposed to minimize the energy
consumption, makespan, and task rejection rate of the cloud system. The proposed algorithm efficiently allocate tasks to VMs and then VMs to hosts. The performance of the proposed algorithm, i.e., Energy-aware Task-based Virtual Machine Consolidation (ETVMC) and the existing algorithms: First Come First Serve (FCFS), Round-Robin, and EERACC proposed in [31] are compared by varying the number task and number of VM with the help of CloudSim simulator. The simulation results indicate minimum energy consumed by the proposed algorithm in comparison to other existing algorithms

Dynamic Voltage and Frequency Scaling (DVFS) technique is a technique through which energy consumption can be minimized for computing resources. We have proposed a heuristic algorithm, i.e., Energy-Efficient DVFS-based Task Scheduling Algorithm (EEDTSA) for the selection of VM for each task to optimize the energy utilization by applying the DVFS technique. The DVFS Mechanism is applied to the virtual machines level to reduce the energy of the cloud system. Moreover, the performance of the diverse algorithms (Random allocation, and FCFS) are compared with the proposed DVFS-based VM selection algorithm. It can be observed from the simulation results that the proposed algorithm (EEDTSA) offers greater energy saving as compared to other existing techniques.

We have proposed a mobile cloud system with edge data center interfacing mobile user
to the cloud system. There are three computing entities (VMs that runs on the top of the host in the data center, VMs that runs on the top of the edge computing devices, and mobile computing devices) used by the mobile user. An energy efficient task allocation in mobile cloud system scheme, i.e., Energy-Efficient Task Allocation in Mobile Cloud System (EETAMCS) is proposed where the selection of appropriate VM for a task with a deadline is explained. Instead of offloading of tasks directly to the cloud data center, in the proposed scheme, the tasks can be offloaded to the edge data center to minimize the energy consumption and execution delay. The result analysis of the proposed algorithm obtained indicates the utilization of edge data centers reduces energy consumption and execution delay