An Adaptive Virtual Machine Location Selection Mechanism in Distributed Cloud

Weijia Jia, Shukun Liu, Weijia Jia 0001Ksii Trans. Internet Inf. Syst 2015 · DOI

Abstract

The location selection of virtual machines in distributed cloud is difficult because of the physical resource distribution, allocation of multi-dimensional resources, and resource unit cost. In this study, we propose a multi-object virtual machine location selection algorithm (MOVMLSA) based on group information, doubly linked list structure and genetic algorithm. On the basis of the collaboration of multi-dimensional resources, a fitness function is designed using fuzzy logic control parameters, which can be used to optimize search space solutions. In the location selection process, an orderly information code based on group and resource information can be generated by adopting the memory mechanism of biological immune systems. This approach, along with the dominant elite strategy, enables the updating of the population. The tournament selection method is used to optimize the operator mechanisms of the single-point crossover and X-point mutation during the population selection. Such a method can be used to obtain an optimal solution for the rapid location selection of virtual machines. Experimental results show that the proposed algorithm is effective in reducing the number of used physical machines and in improving the resource utilization of physical machines. The algorithm improves the utilization degree of multi-dimensional resource synergy and reduces the comprehensive unit cost of resources.

Introduction

In distributed cloud environments, the efficient placement of virtual machines (VMs) onto physical machines (PMs) is a critical challenge. This task is essentially a multi-dimensional vector packing problem, requiring the simultaneous balancing of CPU, RAM, disk space, and I/O bandwidth. Suboptimal placement leads to either performance bottlenecks due to over-packing or energy waste due to under-utilization. Traditional methods often focus on single-resource dimensions or ignore the operational costs associated with VM migration. This research introduces MOVMLSA to address these gaps by optimizing for global resource utility, energy consumption, and system stability.

Methodology

The Multi-Object Virtual Machine Location Selection Algorithm (MOVMLSA) integrates a Genetic Algorithm (GA) framework with several biological and structural enhancements.

Mathematical Modeling and Notation

To formalize the optimization problem, the researchers defined a comprehensive set of symbols representing the cloud environment's state, including resource sets, utility metrics, and scheduling matrices. These notations are essential for defining the multi-objective fitness function.

Table 1: Mathematical symbols and descriptions for the MOVMLSA model.
Table 2: Continued mathematical symbols and descriptions.

Key Algorithmic Components

  1. Doubly Linked List Encoding: Unlike standard fixed-length arrays, MOVMLSA utilizes a doubly linked list structure for chromosome representation. This facilitates group-based encoding where VMs are clustered by their host physical node, enhancing the efficiency of crossover and mutation operations.
  2. Immune Memory Mechanism: Inspired by biological immune systems, the algorithm maintains a "memory set" of high-performing, non-dominated solutions. By utilizing an $\epsilon$-dominant relation, the algorithm avoids losing optimal placement strategies and accelerates convergence compared to standard GAs.
  3. Multi-Objective Fitness Function: The fitness function evaluates placement strategies based on four primary objectives: minimizing active physical machines, reducing VM migration overhead, maximizing multi-dimensional resource utilization, and minimizing load imbalance variance ($\mu$) across the cluster.

Experimental Evaluation

The performance of MOVMLSA was evaluated using the CloudSim toolkit. The simulation environment consisted of 40 physical nodes, each modeled with dual-core processors, 4GB RAM, and dual 500GB disks. The algorithm was tested against varying VM counts (30 to 80) and compared with three baseline methods: Priority Heuristic (PHM), Non-dominated Sorting Genetic Algorithm II (NSGA-II), and Dominant Resource Fairness (DRF).

Results and Discussion

Experimental data indicates that MOVMLSA consistently outperforms baseline algorithms across key performance indicators:

  • Resource Utilization: At high workloads, MOVMLSA achieved a resource utility rate of 82%, representing a 10–14% improvement over PHM and NSGA-II. This is attributed to the algorithm's ability to exploit "multi-dimensional synergy," pairing VMs with complementary resource requirements.
  • Server Efficiency: In a scenario with 80 VMs, MOVMLSA required only 26 physical machines, whereas the PHM method utilized all 40 available nodes.
  • Load Balancing: The algorithm maintained a significantly lower load balance variance (0.0018) compared to PHM (0.0039), ensuring a more uniform distribution of tasks across active servers.

Conclusion

MOVMLSA demonstrates a robust approach to VM placement in distributed clouds by leveraging immune memory and flexible encoding. While the algorithm significantly improves resource efficiency and load balancing, future work will focus on automating parameter tuning (such as crossover and mutation rates) and exploring the optimal scaling of the immune memory library to further enhance performance in dynamic cloud environments.

Summary updated

BibTeX
@article{jia2015adaptive,
  author = {Weijia Jia and Shukun Liu and Weijia Jia},
  title = {{An Adaptive Virtual Machine Location Selection Mechanism in Distributed Cloud}},
  journal = {Ksii Trans. Internet Inf. Syst},
  year = {2015},
  doi = {10.3837/tiis.2015.12.003},
}
Figure