Research On A Network-aware Virtual Machine Placement Algorithm In Mobile Cloud Computing Environment

Cloud computing is a distributed computing model which provide computingservices and resources on demand. It allows users to rent servers, software, networks,and other IT resources through Internet remotely as real-time computing services.These resources, usually collectively called clouds, are owned and managed by cloudservice providers and can be accessed by end users remotely via the Internet. The pastfew years have witnessed a rapid shift of computing from the desktop to the cloud.With the rapid advancement of both wireless network technologies and mobile smartphones, the demand for mobile devices to run heavier applications is on increase, andthere is an increasing need for cloud services to be provided to mobile users viamobile wireless networks. This new research field is called mobile cloud computing(MCC).The aim of this paper is to contribute towards this young but vibrant field ofMCC by investigating a key problem for cloud computing, i.e., virtual machine (VM)placement. For cloud computing and mobile cloud computing, VMs are the keycomponent of a cloud and they provide the elasticity boasted by cloud computing.Virtual machine has become an essential resource management technique in cloudenvironment. When a new cloud application or service is initiated on a mobile device,a new VM needs to be created to serve this application. On which cloud server toplace this new VM is a key issue and thus the focus of this paper. When the admittedservice gets more demanding on computational resources a VM can be automaticallycreated by the cloud to offload some balance to keep the cloud users satisfied. There ismuch work done on VM placement, mainly focusing on fixed networks. Most of themdevise a function of resource utilizations of individual resource types to decide whereto place a VM.In our paper the network cloud servers run on is wireless network. In particular itis a kind of mesh network where both wireless base stations or access points (APs)exist supporting both point to multi-point communications and ad hoc point to pointcommunication. In this paper we consider a cloud computing environment where both storage resource and computation resource exist. Namely there are both storage cloudand computation cloud and they are physically separate and interconnected viawireless networks. In particular storage cloud is deployed next to the wireless APs andis connected to the APs using wireline network. In contrast, computation clouds aredeployed on end user’s mobile devices. Many cloud applications are data-sensitive,which need to process a large amount of data. Cloud applications process these datathrough the use of a large number of virtual machines, while the total completion timeis an important performance metric. Due to instability of wireless links and thedistance and thus channel quality and data rate between the mobile device where theVM is to run and the AP the data is stored next to, it is essential to consider networkfeatures of this link when choosing a cloud server to place this VM. The mostimportant feature to be considered is the link bandwidth as it is also the majorimpacting factor for delay from compute nodes to data nodes.By taking into consideration network features of a wireless mesh networkenvironment where MCC system is deployed, this paper proposes an effectivenetwork-aware VM placement algorithm. This algorithm tends to place a newlycreated VM for a cloud service to a cloud server, which itself is also a mobile device,that occurs lowest network delay when accessing to application data stored in wirelinenetworks. The aim is to reduce cloud service response time as much as possible inorder to deliver better quality of experience to cloud end users over more noisy andless robust wireless networks. Considering the resource-constrained mobile devices,aiming to use CPU, memory, etc. resources more balanced, this paper furtherproposed a multi-objective optimization VM placement algorithm on basis of thealgorithm. The latter algorithm is a compromise solution which minimizes networklatency and balances use of resources by introducing The Technique for OrderPreference by Similarity to an Ideal Solution (TOPSIS) method.In summary, the major contributions of this paper lie in the following severalaspects. Firstly, the proposed deployment of cloud storage and cloud computationalresources onto wireless networks is relatively new, which considers bothinfrastructure-based and ad hoc network architecture and is sensible to the nature ofcloud data storage and cloud computation resource respectively. Secondly, theproposed network-aware VM placement algorithm effectively works on the abovenew MCC network environment and is efficient in terms of cloud service response time, especially suitable for data-intensive applications. The other proposednetwork-aware VM placement algorithm based on TOPSIS method whichcomprehensively considers the network factors and the target of a balanced use ofresources, can achieve shorter service response time and more balanced use ofresources. It is an effective and adaptable compromise solution to solvemulti-objective optimization. Thirdly, we propose a dual-threshold network-awareVM migration strategy for solving VM reallocation dynamically when networkconditions change or the power of a mobile device becomes low. In addition, weintroduce the autoregressive model of time series forecasting method to avoidfrequent VM migration and reduce overhead of unnecessary VM migration. Thesimulation results have shown the effectiveness and the efficiency of the proposedVM placement algorithms and VM migration strategy both in terms of cloud serviceresponse time, etc.