A Study On Virtual Machine Scheduling In Cloud Data Centers

Cloud computing has gained a wide and highly attention of the industry sector and academic society in recent years. As one of the key technologies in cloud computing, virtualization is seen as one of the main means for resource management. Different mappings between virtual machines (VMs) and physical machines (PMs) cause various resource utilization, which is seen as the issue of virtual machine scheduling. Many VM scheduling schemes propose to reduce the resource energy consumption, improve application performance or provide fault tolerance, but few of them concern the impacts and effects of different scheduling between virtual machines and physical machines on the shared networks, which will undoubtedly affect the application performance in the data, center. Thus, this study focuses on the impacts of virtual machine scheduling on the networks in the data center. The issues of energy optimization and network performance are typically concerned. On the premise of ensuring tenants’ requirements and network performance, how to develop reasonable strategies to reduce the energy consumption and improve the resource utilization is becoming a critical issue for cloud providers.This study firstly conducts a systematic research on the background and the latest developments of virtualization, and gives concrete analysis on the current virtual machine scheduling. With the final objectives of reducing the energy consumption and optimizing the network performance, considering the characteristics of the network topology and traffic in the cloud data center, this study focuses on the issue of virtual machine scheduling mechanism in order to optimize the resource efficiency. Four schemes are included in this study:virtual machine placement scheme for optimizing server and network resources, virtual machine placement scheme for improving the distribution of network traffic, virtual machine migration scheme for balanced optimization between network performance and energy consumption, and lastly, virtual machines dynamic migration scheme for reducing the network congestion. The details are as following:1. Virtual machine placement for optimizing server and network energy consumption. In IaaS Cloud, different mapping relationships between virtual machines (VMs) and physical machines (PMs) cause different resource utilization, so how to place VMs on PMs to reduce energy consumption is becoming one of the major concerns for cloud providers. The existing VM scheduling schemes propose to optimize PMs or network resources utilization, but few of them attempt to improve the energy efficiency of these two resources simultaneously. This study proposes a scheme to simultaneously optimize physical servers and network resources (such as switches, routers, network links, etc.), and also proposes to close the sleeping and idle servers, switches, network links and other physical resources so as to finally reduce the energy consumption. This virtual machine placement is abstracted as a multi-dimensional packing and traffic routing optimization issue, which is a classic combinatorial optimization as well as NP-hard problem. Correspondingly, an ant colony optimization algorithm is designed to solve this issue, and the simulations show that our solution achieves good results.2. Virtual machine placement for improving the distribution of network traffic. With the wide application of virtualization technology in cloud data center, how to effectively place virtual machine (VM) is becoming a major issue for cloud providers. The current virtual machine placement (VMP) solutions are mainly to optimize server resources, but network resources optimization is less concerned, and the impacts of the network topology and the current network traffic are also ignored. Thus, this study proposes a virtual machine placement scheme for simultaneously optimizing network traffic and link utilization so as to maintain a balanced traffic distribution and avoid the congested links, which is a classic combinatorial optimization and NP-hard problem. Ant colony optimization and local search are combined to solve this problem, and the simulation results show that MLU is decreased by20%, and the number of hot links is decreased by37%by our combined scheme.3. Virtual machine scheduling for balanced optimization between network performance and energy consumption. How to design a scheme to reduce energy consumption while improving the network performance is becoming an important issue for cloud providers in the cloud data center. Current VM scheduling schemes are mostly to reduce energy consumption by optimizing the utilization of physical server or network elements. However, the aggressive consolidation of these resources may lead to network performance degradation. In view of this, this study proposes a two-stage VM scheduling scheme:(1) static VM placement is abstracted as multi-dimensional constraint packing and quadratic assignment problem, and the numbers of activated physical resources such as physical servers, switches, and other network links are minimized to realize the objective of reducing energy consumption;(2) on the premise of minimizing the migration costs, a dynamic VM migration scheme is proposed to minimize the maximum link utilization in order to improve the network performance. This scheme realizes the aim of optimizing network performance while reducing the energy consumption of physical servers and network devices, making a tradeoff between energy efficiency and network performance. Correspondingly, a new two-stage heuristic algorithm is designed, and the simulations show that our solution achieves better results.4. Virtual machines migration scheme for reducing the network congestion. Network congestion frequently occurs in the current cloud data centers, leading to the increased packet loss, latency and degrading throughput. Thus, the network has become a performance bottleneck and network congestion needs to be solved urgently. Rescheduling flow path, modifying the network protocol stack and other methods are traditionally applied to solve the network congestion, but this study attempts to apply virtual machine migration to change the position of flow sender or receiver in the network topology, which can effectively reduce the network congestion and optimize traffic layout. Compared with the methods of modifying the flow path or protocol stack, our proposal takes advantage of the elastic features of cloud service, so our proposal achieves better results with its simplicity. Two aspects are considered in this migration scheme:(1) to minimize the traffic costs between virtual machines;(2) to minimize the migration cost. In view of this, our scheme proposes a three-stage virtual machine migration based on greedy algorithms, which can achieve better time costs. The simulation shows that our algorithms effectively reduce the network congestion on the premise of minimizing the migration costs.