Article Preview
Top1. Introduction And Motivation
Cloud computing is one of the new forms of data processing or storage via a network. The objective is to “provide on-demand, secure, qualitative, scalable, fast, more responsive, profitable, and automatically provisioned services based on the pay-as-you-go model” (Zhou et al., 2020). Software as a Service (SaaS), Platform as a Service (PaaS), and Infrastructure as a Service (IaaS) are the main three service models, recognized by the National Institute of Standards and Technology (NIST). In addition, there are other well-known services such as Database as a Service (DbaaS), Container as a Service (CaaS), etc. (Helali & Omri, 2021). Every cloud service provider needs to store their data in datacenters. The latter consumes a lot of energy (El Kafhali & Salah, 2018). This energy consumed by the infrastructure results in significant production of carbon dioxide emissions (CO2). In the literature, it has been revealed that a host uses 10-50% on average of its total capacity, and data centers consume approximately 3% of the total electricity (Saadi & El Kafhali, 2020), which leads to problems related to power consumption, energy costs, and performance. There are other components related to this consumption, such as processing, network, cooling systems, and disk storage (Abdessamia et al., 2017). The most relevant technology addressing this problem in the infrastructure part is virtualization, which allows cloud providers to set many VM instances on a single PM. Virtualization targets resource sharing in cloud systems. This makes it possible for the applications to execute on different platforms, which allows a reduced number of VMs executed in PMs (Hanini & El Kafhali, 2017). This process is named consolidation. Dynamic VM consolidation uses live migration to improve energy efficiency for cloud computing providers without wasting resources with respect to QoS (Ding et al., 2020). Consolidation techniques involve several features, such as objectives, workload types, or optimization methods for VM placement. The main goal of these techniques is to improve resource management in such a way as to reduce energy consumption and to meet QoS requirements.
The consolidation process can be split into three main steps, each has its own policy. The first policy allows the detection of PM load, the second one selects VMs that should migrate to another PM, and the last one searches for a new PM for the migrated VM.
Due to its complexity, VM consolidation has typically been split into several smaller issues to reduce energy usage at data centers in an efficient manner. Moreover, numerous survey papers present various themes on server consolidation. For instance, energy-aware server consolidation (Chaurasia et al., 2021) and VM consolidation review (Zolfaghari et al., 2021) are based on various criteria: metrics, migration techniques, objective functions, co-location for VMs, evaluation techniques, algorithmic techniques, and workload datasets. Diverse contributions in server consolidation, VM migration, and DVFS schemes were studied in (Shirvani et al., 2020). Large-scale datacenters' energy usage, performance, and cost challenges are discussed in (Khan & Zakarya, 2021). The authors in (Wang et al., 2020) developed a taxonomy to research and categorize workload scheduling, and resource provisioning in hybrid cloud systems. The concepts of virtualized data centers and consolidation in cloud computing systems are described in (Helali & Omri, 2021) holistically. Finally, the work (Alashaikh et al., 2021) presents an analysis of preferences in the most recent research on VM placement.