Self-Management of Resources on Container based Infrastructure for Desktop-as-a-Service: A Case Study on Computer Laboratories of ECT/UFRN
Self-adaptive software, Container-based Virtualization, Containerization, LTSP, DaaS, Desktop-as-a-service
A viable alternative for institutions that have multiple users who need access to desktop applications is Desktop-as-a-Service (DaaS), which is characterized by the delivery of a desktop environment that runs remotely. Resource virtualization and load balancing are widely used techniques in infrastructures that host services with seasonal demands, replicating instances and distributing requests among them to achieve elasticity. However load balancing is not the most suitable solution for DaaS, since sessions in this service are long lasting and are not migrated to a new server that is added to the balancer, remaining the slowness perceived by users already connected to an overloaded server. In this context, the dynamic resizing of resources in a virtual instance is shown as the most appropriate approach. However, traditional virtualization solutions require a reboot of the affected server, and consequently, terminating DaaS sessions with their respective users. On the other hand, container-based virtualization allows such resizing, but requires manual administrator intervention to adjust the amount of resources on demand. This work presents ConManager, a self-adaptive controller for container-based environments, which aims to dynamically resize virtualized resources to handle seasonal loads. The proposal has been applied as a case study in the computer laboratories of the Escola de Ciências e Tecnologia of the Universidade Federal do Rio Grande do Norte. ConManager monitors the use of resources in laboratories, detecting overhead scenarios, and proposing adaptation plans that are applied to the DaaS service support infrastructure, effectively redistributing resources from underutilized containers to overloaded ones. The tool is currently in use and has brought noticeable gains such as reduced time to adapt resources and simplified environmental management, benefiting the institution's information technology team, responsible for maintaining the service and the academic community that enjoys a Stable computing environment.