Numerous studies are presented in the area of cloud scheduling, and some generic challenges are discussed such as resource scheduling, resource provisioning, and load balancing. Extensive surveys have also been found in the literature on virtual machine scheduling policy - VM placement, VM allocation, VM migration and VM scheduling. However, there is no extensive systematic survey on virtual machine scheduling in current studies. This section refers to some studies in the area of cloud scheduling. When it comes to allocating dynamic, heterogeneous, and shared resources, resource scheduling in cloud environments is considered to be one of the most crucial challenges. To provide reliable and cost-effective access, overloading of those resources must be prevented by proper load balancing and effective scheduling techniques.
Detailed review and classification of load balancing techniques are presented in [16], in which they compare the existing state-of-art techniques on parameters such as model, strength, gap, techniques and future work. Moreover, they have sufficiently analyzed and presented job migration techniques considering their description, merit, and demerits, which play a vital role in achieving fault tolerance. However, the study has not considered some of the job migration studies like predictive and heterogeneous job migration. At the same time, the scope of the study remains within grid computing.
In [17], a comprehensive survey of cloud scheduling algorithms which offer an analysis based on the categorization of some parameters that include; load balancing, energy management, makespan, and many more. The study observed that there is any scheduling algorithm that has the potential to effectively address all parameters of VM scheduling. Furthermore, the study discussed some task scheduling algorithms, limitations, and some future problems. However, the scope of the study is restricted to only grid computing. Similar work in [18, 19] presented a study of scheduling and energy-conscious resource allocation methods with a focus on the quality of service. They mentioned some critical and open challenges in cloud scheduling, particularly energy management in a cloud datacenter. According to their analysis of previous studies, the challenges are enumerated as follows: (1) Processes that are quick and energy-efficient for placing virtual machines and can anticipate workload peaks to prevent performance deprivation in a heterogeneous environment (2) energy-based virtual network topology optimization technique amongst VMs for the best location to lessen network traffic congestion, (3) to properly regulate temperature and energy use, new heat management algorithms, (4) even workloads and workload-aware resource allocation processes, and (5) Scalability and fault-tolerance techniques for virtual machine placement (VMP) challenges that are decentralized and distributed.
Virtual machine migration is a major issue for scheduling. In [20], the paper analyzed current VM migration techniques of thematic taxonomy that underline the commonalities and variances among VM migration schemes concerning certain performance metrics.. Additionally, they look into the difficulties with the VM migration plan, including the heterogeneity of cloud resources, the nature of dynamic workloads, system burden, VM memory size, and the severity of SLA breaches. Considering security is one of the significant concerns in the VM migration process, they suggest some safeguards such as (1) stopping unauthorized parties from accessing VMM; (2) separating VM borders; and (3) network connection security [21].
In another study, [22] investigated live VM migration schemes and present a particular taxonomy to categorize the concerned literature. They investigate storage optimization methods for WAN links, server consolidation rules, DVFS-enabled power optimization, and bandwidth optimization techniques based on their categorization. They also give a comparison of the results of other polls, highlighting some of the crucial factors in virtual machine migration. Their investigation identifies similarities and contrasts across existing VM migration plans based on a set of parameters found in the literature. Their research may be useful for creating intricate designs and optimization strategies for VM migration methods. The mathematical modeling of virtual machine migration strategies, however, is lacking in this research.
In a similar kind of work, Li, Li [23] investigated the scheduling issue for virtual machines in a cloud data center. Additionally, they provide a survey of current technologies, including virtualization, resource scheduling, virtual machine migration, security, and performance assessment in cloud computing. Similar to this, they cover certain upcoming problems and difficulties such as CPU architecture, resource management, upkeep procedures for system security, and performance assessment techniques in a system with several virtual machines. However, the paper lacked categorization, issue formulation, and parametric analysis, and did not conduct a thorough investigation of the methodologies as indicated in earlier research.
Analyzing the cloud computing architecture, Zhan, Liu [24] systematically presented two-level taxonomy of cloud resources. Researchers have critically examined the issue and remedy of cloud scheduling in their review. Additionally, they investigated EC methodologies and talked about several cutting-edge evolutionary algorithms and their potential to solve the cloud scheduling issue. Based on their categorization, they have also identified some of the next problems and research fields, such as distributed parallel scheduling, adaptive dynamic scheduling, large-scale scheduling, and multi-objective scheduling. They have also highlighted some of the most cutting-edge future themes, including the Internet of Things and the convergence of cyber and physical systems with big data. However, they failed to describe the problem's mathematical modeling or include any parametric analysis in the paper.
In another investigation, Xu, Liu [25] described the causes of the performance overhead problem of a virtual machine under several scenarios i.e., from single server-virtualization/ datacenter to multiple and distributed datacenters. The review presents a detailed comparison of contemporary migration techniques and modeling approaches to manage performance overhead problems. However, the authors suggest that there remains a lot to be resolved to ensure the predictable performance of VMs with guaranteed SLA. Similarly, Madni, Latiff [26] examine the difficulties and possibilities in resource scheduling for cloud infrastructure as a service (IaaS). They categorize the previous scheduling schemes according to the issues addressed and performance metrics and present a classification scheme. Furthermore, some essential parameters are evaluated and their strengths and weaknesses are highlighted. Finally, they suggest some innovative ideas for future enhancements in resource scheduling techniques.
One of the significant and recent studies suggesting a taxonomy of the algorithms for load balancing in virtual machine placement is presented by [27] The work is based on several existing models and techniques of load balancing algorithms employed in the virtual machine placement method. Their work summarizes various parameters and their optimization, contribution, gap, future challenges, and improvements. However, the scope of the review is limited to the virtual machine placement problem and ignores VM migration.
Meta-heuristic techniques become a benchmark in cloud computing scheduling because they exhibit efficient and near-to-optimal results in a reasonable time-space. Several types of research have been carried out to assess how well these modern meta-heuristic algorithms perform. In a similar study, Kalra and Singh [28] investigated six major Mete-heuristic optimization techniques namely, Ant Colony Optimization (ACO), Genetic Algorithm (GA), Particle Swarm Optimization (PSO), League Championship Algorithm (LCA) and Bat algorithm. Each Mete-heuristic technique is described in a taxonomical framework, and each technique is compared using some scheduling criteria, such as task awareness, SLA awareness, and energy awareness. Moreover, they have discussed the application of these meta-heuristic techniques and open challenges in the area of grid or cloud scheduling. However, the survey is only limited to specific meta-heuristic techniques and optimization criteria.
In another development, Madni, Latiff [29] investigated the potential of existing state-of-the-art Mete-heuristic techniques for resource allocation in a cloud computing environment for maximizing financial benefit for the cloud provider and minimizing cost for cloud users. In their research, they selected 23 meta-heuristic technique studies between 1954 and 2015. They compared meta-heuristic techniques with traditional techniques to evaluate the performance criteria of the algorithms. They claim that there can be several ways to enhance the performance of these algorithms which can further solve the resource scheduling problem. However, this review resembles the work of [28]. However, the focus of the paper is only on meta-heuristic methods.
Unlike previous studies shown in Table 1, our research presents an extensive (not exhaustive) review of virtual machine scheduling techniques and presents the most appropriate categorization, problem formulation, architecture and future challenges. Then, based on our research, we formalize three questions and choose the most important study from the most trustworthy research database to address them. Furthermore, we delineate the importance of virtual machine scheduling techniques, current issues and challenges, and future direction to support future research.
Table 1
Summary of previous literature in virtual machine scheduling
Previous Reviews | VM Scheduling | Problem formulation | Classification of VM Scheduling | Parametric Analysis | Simulation Tool & Environment | Dataset Available | Architecture | Period Covered |
---|
Li et al. [54] | | | | | | | √ | 2002–2009 |
Beloglazov et al. [9] | √ | √ | | √ | | √ | √ | 1991–2012 |
Rathore and Chana [50] | | | | √ | | | | 1999–2014 |
Xu et al.[56] | | √ | √ | √ | | | √ | 2003–2013 |
Abdulhamid et al. [51] | | | | √ | √ | | √ | 2009–2014 |
Kalra and Singh [58] | | | √ | √ | | | | 2001–2005 |
Zhan et al.[55] | √ | | √ | | √ | | √ | 2003–2014 |
Ahmad et al.[53] | √ | | √ | √ | √ | | √ | 1993–2014 |
Ahmad et al. [52] | √ | | √ | √ | | | √ | 1997–2015 |
Madni et al.[59] | | | √ | √ | | | | 1954–2016 |
Madni et al. [57] | | √ | √ | √ | | | √ | 2008–2016 |
Xu et al. [46] | | | √ | √ | | | √ | 2008–2016 |
Our Review | √ | √ | √ | √ | √ | √ | √ | 2008–2022 |