Energy efficient resource allocation of cloud data centers mainly focuses on maximizes CPU utilization, minimizeresource usage and reduce carbon emission extensively. Moreover, to reduce energy consumption DVFS (Dynamic VoltageFrequency Scaling) technique places significant role in all electronic devices such as, desktop, laptop, mobile phones and allhandheld devices.
It has in-built frequency controller to supply appropriate voltage based on the requirement of VMs. FeiCao et al. 4 have proposed an energy efficient workflow scheduling based on Directed Acyclic Graph (DAG) model. In thispaper, DVFS technique applied for resource provision and allocation of cloud data centers. In addition, authors have alsoinvestigated few areas like, resource utilization, high performance, VM overheads, energy consumption and Co2 emissions.However, all approaches works efficiently while compromised SLA violations.
Devendra Singh Thakur 5 has calculated resource usage for single application using single host mechanism. In this,authors have introduced PaaS layer for task scheduling instead of IaaS layer. In addition to that, heterogeneous workloadconsolidation technique used to find energy consumption of each data centers. It also includes allocation policies of PaaSlayer with SLA; restrict the users from overloaded resource usage. However, it works only on single application mechanismbut not in multi-platform or cross-platform computing software.
Ning Liu et al. 6 have studied about optimization model fortask scheduling. It is used to minimize energy consumption using integer programming.
Furthermore, Greedy taskscheduling algorithm used to reduce average response time and total number of active servers needed to execute a task. Inorder to makes an efficient utilization of data centers, Most Efficient Server First Task Scheduling (MESF) algorithm used tofind energy usage cost. Finally, the result shows that 70% energy saved. Meanwhile, authors mainly focused on responsetime reduction but not on server load balance and system performance.Hongyang Sun et al. 7 have considered the problem of effective resource management on homogeneous HighPerformance Computing (HPC). In which, it includes spatio and temporal thermal aware scheduling for temperature awareenergy saving methodology. Temperature progress obtained using thermal model like spatial and temporal correlations.
Furthermore, when minimize makespan of the HPC system; authors have validated online heuristic scheduling algorithm.This algorithm performs both job scheduling and thermal management system. However, proposed model focuses only onhomogeneous HPC system but not on heterogeneous cloud environment. Y. Peng et al.
8 have discussed about Energy andQoS to minimize performance degradation and bandwidth cost. Based on these two constraints, authors have proposed a newalgorithm called Evolutionary Energy Efficient Virtual Machine Allocation (EEE-VMA). Moreover, genetic basedmeta-heuristic model, which considered for heterogeneity power aware VM allocation. This technique named as PowerMark,used to monitor power efficiency of cloud data centers.
Despite of efficient power aware scheduling, authors have notconsidered SLA violations.Saurabh Kumar Garg et al. 9 have addressed the problem of HPC. It is not only profitable for commercial applicationsbut it also takes significant role in consumer IT applications. In this paper, authors have introduced a near optimal schedulingpolicy for heterogeneity data centers in geographical locations.
Proposed model formulated based on the data centers location,architectural design and management systems. Like 8, energy consumption and power aware problem discussed withoutconcentrating on service level objectives. Anton Beloglazov et al. 10 have proposed optimizing resource usage and energyconsumption in cloud data centers.
It is achieved by dynamic virtual machine consolidation techniques using live migrationand turning idle machines to sleep mode. In addition, authors have addressed online deterministic algorithm used for singlehost migration and dynamic VM consolidation. Furthermore, based on historical data, dynamic consolidation and resourceusage were calculated. However, single host migration inefficient while compare with multi host consolidation techniques.Jing Liu et al.
11 have investigated task-scheduling model based on Multi-Objective Genetic Algorithm (MO-GA). Inthis algorithm, the following parameters gets involved like, encoding rules, crossover operators, selection operators andmethod for sorting pareto solution. However, MO-GA is suitable for efficient resource usage in cloud environment but notconsidering on strict QoS and SLA requirements. XiaochengLiu et al. 12 have suggested two-tier VM architecture namedas Aggressive Consolidation based FCFS (ACFCFS) for parallel workload consolidation.
Experimental results comparedwith traditional FCFS, EASY (Extensible Argonne Scheduling sYstem), CCFCFS (Conservative Consolidation basedFCFS). The proposed algorithm outperforms, and it allows inaccurate CPU usage for parallel job execution. Furthermore, itreduces available CPU idle time and starvation problem. Limitations of ACFCFS algorithm do not satisfy energyconsumption of cloud VMs.Anton Beloglazov et al. 13 have discussed green cloud architectural framework for energy efficient computing.
Themain objective of this paper addressed energy aware heuristic resource scheduling and VM allocation policies. It isexclusively used for cost saving and reducing energy consumption of allocated VMs. Moreover, authors have proposed twonew algorithms called, MBFD and Minimization of Migration (MM) policy for efficient utilization of cloud resources.Therefore, exploration of these two algorithms works efficiently with negotiated QoS. However, the authors have defined anefficient mechanism without focusing on dynamic requirements of cloud resources. Chia-Ming Wu et al. 14 have pointedout green cloud is an emerging technology and it increases resource utilization and reduce high power consumption in clouddata centers.
In this paper, authors have proposed green cloud scheduling algorithms using DVFS technique. Furthermore,priority allocated based on VM weights to assign resources in data centers. Meanwhile, decreasing energy consumption andexecution time authors have not concentrated on load balancing when overloaded resource used.
Fabio D. Rossi et al. 15 have proposed an orchestration of different energy saving techniques to improve energyconsumption and application performance. Energy Efficient Cloud Orchestrator defined as E-eco helps to reduce energyconsumption in dynamic cloud environment. They have proposed an orchestration E-eco compared with power-agnosticapproach.
However, performance-aware approach sacrificed 6% of SLA violation and it creates an impact of applicationperformance degradation. Yuyang Peng et al. 16 have introduced an EEE-VMA for geographical cloud data centerslocations.
It works based on Genetic Algorithm (GA) metaheuristic optimization technique. The proposed model supportspower heterogeneity aware VM allocation using powerMark technique, which diagnoses the power efficiency of differentcloud data centers. Furthermore, they have investigated performance degradation due to VM co-location and bandwidth costbetween cloud users and providers. Limitations of this proposed approach immolate QoS requirements.Ehsan Arianyan et al. 17 have investigated energy consumption using consolidation of VMs in cloud data centers.
Inthis paper, authors have proposed holistic resource management and heuristic based multi-criteria decision-making method.It is used to discover underloaded host then move to the appropriate server. The objective of this paper, which succeeds inenergy consumption, SLA violation and reduction of VM migrations, but it failed to achieve performance of the system.Jiyuan Shi et al.
18 have defined the problem of efficient resource allocation for large-scale data centers. Optimizationproblem arises when multi-dimensional resources requested from cloud users. To solve this problem authors have proposedpattern based resource allocation mechanism for efficient utilization of server resources. To reduce number of running VMs,they have considered pattern based online and offline VM scheduler techniques. However, authors not focused on QoS andapplication performance.