Research On Task Scheduling In Virtualized Environments

Virtualization is an essential technology in modern datacenters which is becoming more and more popular. By running multiple virtual machines in a shared physical machine, virtualization enables high utilization of hardware resources. Large-scale data centers leverage virtualization technology to achieve excellent resource utilization, scalability, and high availability. Different with the traditional ones, data centers with virtualization which offer resources as virtual machines have the characteristics of dynamic, openness and heterogeneous. The user groups of data centers is the general public, so its user community is very extensive and its amount stay handle tasks is huge;System almost always manage a large number of computing tasks, so we need an efficient algorithm for task scheduling strategy in the cloud environment. Besides, according to the actual state of the virtual environments to achieve comfortable resource assignment and task scheduling can be key technology for minimizing schedule length, high reliability and efficient energy consumption.This paper has done a research in task scheduling in the virtualized environments. Based on previous researches, major research works and innovative points in this paper are:1. This paper deeply analyzes the characteristic of virtualized environments based on virtualization technology and uses graph technology to construct virtualization characteristics models. By deeply analyzing the characteristic of virtualized environments, the paper constructs the virtualization computing model, virtualized resource scheduling model and task (independent task, dependent task, real-time task) model.2. This paper proposes a list scheduling algorithm named ELS which concerning both completion time and energy consumption for independent tasks with completion time constraint based on virtualized characteristics model. We prove in Propositionl that the slower speed (MIPS) of the virtual machine that the task assigned to, the less energy consumption of the task. Based on Propositionl, ELS tries to assign tasks to the virtual machine with low speed with the completion time threshold to reduce the energy consumption of tasks. The experiments show that ELS could significantly minimize the energy consumption with the completion time constraint.3. This paper establishes united cost function of schedule length and reliability for independent tasks. Based on the united cost function, the paper proposes a hybrid genetic algorithm that concerning both schedule length and reliability. The proposed algorithm firstly runs the Min-min algorithm to generate initial solutions. Then the algorithm runs the genetic algorithm with the generated initial solutions to optimize the united cost function to construct efficient task scheduling. The simulation experiments show that the hybrid genetic algorithm outperforms the current scheduling algorithms in terms of schedule length and reliability.