اسم الباحث :
Rana Al-shami Ibrahim Ahmed Al-Baltah
Abstract Due to the dynamic nature of multi-tenant in the cloud environments that provides many available services to the Internet users on demand through virtual technology. Resource Management such as virtual machine placement provides immediate use while ensuring confidence and control over all virtual machines becomes more complex. Many research studies have attempted to improve the self-managed virtual machine placement. However, there are still some critical issues hinder the achievement of the required trust of virtual machines with respect to their protection against the threats of co-residency. Wherefore in this paper we propose a trust model that enhances the reliability of the Self-managed virtual machine placement scheduling through the monitoring of the behavior of devices and the fingerprint extraction of behavior. Keywords: Cloud computing, Virtual machine, Cloud service providers, Trust model, Threats of co-residency With the evolution of cloud computing in recent years, computing has become one of the most widely used systems to provide services such as applications, software, CPU, memory and disk on demand for Internet users. Cloud computing has been referred to as a large-scale distributed computing paradigm [1]. Virtual technology has worked to achieve the goal of cloud computing in the possibility of sharing resources between heterogeneous environments with many users. Moreover it provides a means of improving the use of resources and reduces cost and energy consumption. Commonly user requests from resources on virtual machines, each device consists of a limited quantity of resources (CPU, memory and disk resource). One of the most important problems that service providers should deal with when they are targeting to use cloud computing is to provide the request for resources and place the virtual machine VM in the right place on the physical devices (original resources). This process is called VM placement. The typical placement algorithm should takes care of choosing the best place in physical devices, rack and host within the data center, and most importantly is to make a placement decision that provides sufficient space to accept future resource requests [3]. However, the participation of virtual machines and the same physical devices may expose the system to co-residency threats [4]. Of these threats, unauthorized use, exploitation of resources or information theft [5]. Simple solutions were depend on fixed criteria such as the selection of physical devices that have resources available more than others or some constraints associated with the user position. Other modern solutions include virtual machines placement scheduling in a dynamic manner built on the basis of self-management, which are based not only on fixed criteria, however also on knowledge of what has been implemented in the system over time according to resource monitoring in order to improve efficiency of performance and usage while reducing energy consumption. The authors of [6] referred to real-time resource analysis based on the use of past resources through the development of an automated learning model. This learning model analyzes the usage of both virtual machine and physical machine resources on-the-fly. Nonetheless, these solutions are focused on improving VM placement without addressing the security aspects related to VM placement. Therefore, we propose a model to enhance the reliability of scheduling devices which include the integrity of the virtual machines from threats of co-residency with maximizing the optimization of resource utilization and reduction of performance degradation (performance upgrade). Our contribution is to build a properly monitored model that monitors the behavior of virtual machines and the fingerprint extraction of behavior to generate the machine learning models which can be used in three dimensions. The first is to enhance trust by verify the credibility of current behavior. The second is to speed up the selection process of physical devices by reducing the machine learning models education used to enhance the learning phase from the historical records of using resources in real time. The third is to reduce the storage space of past records where they will only contain the trusted models. This paper is organized as follows. Section (2) presents background; Section (3) reviews the related work; Section (4) presents the proposed model; and finally, we draw some conclusion and future work in section (5).
سنة النشر :
2017
ملخص البحث :
