A Data-driven Resource Allocation Method for Personalized Container based Desktop as a Service

Cloud based virtualization technology gained popularity coupled with sustained growth of cloud computing. DaaS (Desktop as a Service) is a desktop virtualization technology which enables the most powerful service in the cloud environment. It is used in many areas such as financial services, manufacturing, healthcare, and education. Also, it enables fully personalized desktops for each user by providing all the security and simplicity of centralized management. However, most of the existing DaaS technologies are hypervisor based systems. It has a performance degradation problem to loading each desktop image because of the multilayered architecture by the Guest OS and slow creation time of each VM. In addition, it cannot provide optimized resources to personalized services which have different resource demands, as it allocates static resources for CPU, RAM, and Network bandwidth in each VM. In these problems, the DaaS in cloud cannot offer better user experience than local PC environment. In this paper, a data-driven resource allocation method for the container based DaaS system is proposed to solve problem of conventional hypervisor based DaaS and its static allocation. To propose a novel resource allocation method, we perform comparative analysis and simulation according to resource usage and workload. By doing this, the proposed scheme looks forward to accelerate growth of the DaaS through improving the user experience and the resource efficiency of the datacenter by allocating the fine-grained resource.