Virtual Machine Consolidation with Minimization of Migration Thrashing for Cloud Data Centers

Abstract

Cloud data centers consume huge amount of electrical energy bringing about in high operating costs and carbon dioxide emissions. Virtual machine (VM) consolidation utilizes live migration of virtual machines (VMs) to transfer a VM among physical servers in order to improve the utilization of resources and energy efficiency in cloud data centers. Most of the current VM consolidation approaches tend to aggressive-migrate for some types of applications such as large capacity application such as speech recognition, image processing, and decision support systems. These approaches generate a high migration thrashing because VMs are consolidated to servers according to VM’s instant resource usage without considering their overall and long-term utilization. The proposed approach, dynamic consolidation with minimization of migration thrashing (DCMMT) which prioritizes VM with high capacity, significantly reduces migration thrashing and the number of migrations to ensure service-level agreement (SLA) since it keeps VMs likely to suffer from migration thrashing in the same physical servers instead of migrating. We have performed experiments using real workload traces compared to existing aggressive-migration-based solutions; through simulations, we show that our approach improves migration thrashing metric by about 28%, number of migrations metric by about 21%, and SLAV metric by about 19%.

Optimization of Data and Energy Migrations in Mini Data Centers for Carbon-Neutral Computing

“…However, given the dynamic nature of the problem, most approaches are heuristic algorithms. Among those, [34], [35], [36] developed algorithms based on classical bin-packing heuristics, e.g., first-fit decreasing and best-fit decreasing. Meanwhile, as mentioned in the previous section, [22], [23] investigated evolutionary algorithms.…”

“…Each node i ∈ N has R M i MB of RAM, R D i GB of disk storage, a network bandwidth of R B i Mb/s, and CPU load capacity R C i . Please note that instead of representing the CPU resource capacity in MIPS (million instructions per second), as in [34], [36], we define an utilization ratio in the interval [0.0, 1.0], where 0.0 means that the machine is idle and 1.0 means that the CPU is 100% utilized. Furthermore, for the sake of simplicity, this utilization measure is not applied in a per-core basis, but for the whole processing unit.…”

Improved beetle swarm optimization algorithm for energy efficient virtual machine consolidation on cloud environment

“…Randomly initialize the solution (refer to Figure 3) Evaluate the fitness of each solution using Equation (10) Set cycle 1 Repeat While (t < t max ) for i = 1:N update the position X t+1 j using Equation (11) update the velocity V t+1 j using Equation (12) end for for i = 1:N update the right ( X t+1 rs ) and left ( X t+1 ls ) antenna position using Equation (18) update the incremental function 𝜉 t+1 js using Equation ( 16) update the beetle speed Q t+1 js using Equation ( 14) update the next position of the beetle X t+1 js using Equation ( 13…”

“…To manage the above issues, this work proposes another calculation for energy-efficient VM integration. 10 The remaining section of the article is sorted as pursues: Section 2 provides a survey of recent articles that presented VM migration and consolidation in a cloud environment. Section 3 defines the multi-objective functions and Section 4 depicts the proposed system model.…”

“…(3) Where should VMs be placed? To manage the above issues, this work proposes another calculation for energy‐efficient VM integration 10 …”