Optimization enabled deep learning methods for Container-based Cloud Computing Environment

Cloud applications are now evolving towards a more granular micro-service paradigm, where fewer and simpler working elements are joined for providing end-to-end services, in response to the demand for agile development and administration. The increase in the utilization of container have gained more insight and helped in ensuring maximized and large portability with minimized overhead and rapid deployment in the cloud platform. But the rapid growth of container technology has introduced phenomenal changes in the management and automation of containers in the cloud computing environment. In specific, container resource allocation is the most potential challenge realized from the dimension of cloud providers since it possesses a direct impact on system performance and resource management. In this paper, a comprehensive review on Optimization enabled deep learning methods for Container-based Cloud Computing Environment is presented with its merits and limitations. It has presented the possible number of deep learning models-based container scheduling process that helps in significant load prediction in the cloud platform. This study also outlines the prospective research areas that might be explored moving forward with this implementation research.

I. INTRODUCTİON

Containerization puts everything required to operate a single application or micro-service into a single unit rather than requiring the creation of a full virtual machine (along with runtime libraries they need to run). A hypervisor virtualizes physical hardware in classical virtualization. As a result, each virtual machine includes a guest OS, a virtual replica of the hardware that the OS needs in order to function, an application, as well as the libraries and dependencies that go along with it. Containers virtualize the operating system as opposed to the underlying hardware, thus each container just includes the programme and its libraries and dependencies. Unlike virtual machines, which must always incorporate a guest OS, containers can just use the host OS, making them tiny, quick, and portable. In the recent decade, cloud computing which facilitates the required computer services through the Internet has emerged as one of the most potential paradigms [1]. The rapid and phenomenal development of several sophisticated clouds that serves the domains of Things (IoT) devices have evolved due to the recent growth in terms of user demands. It opens the option of provisioning the available services, the memory of the cloud and machine learning programmes which is mandate for possible amenities of the clouds which have rapidly grown in a more substantial manner [2]. In specific, the business related to cloud computing has evolved rapidly in the recent years due to the introduction of various technologies of virtualization that includes Zen, KVM, Citrix and VMware. But the use of widespread technologies of virtualization has brought in about some of the limitations that include procedures of migration, difficult planning, extended shutdowns, extended executions and high time consumption. On the other hand, the virtualization of hardware in the traditional setup, and each individual virtual machine executing the entire operating system is responsible for supervising the entire activities of the computers’ applications [3]. The container never possesses its own virtualization of hardware or kernel, and this process of application directly communicates with kernel of the host in a more reliable manner. Thus the containers are identified to be comparatively lighter than the conventional virtual computers.