Resource Management and Prioritization in an Embedded Linux System

Abstract: This master thesis tackles the problem of limited computing resources on a camera that is executing computing applications together with image acquisition and streaming. The thesis was carried out at Axis Communications in cooperation with the Department of Automatic Control at Lund University. The problem of limited resources on an Axis camera is handled by a two part solution where a resource manager (RM) distributes the available resources and services can adapt their service level (SL) in order to finish their jobs on time. The solution is based on game theory, where services are players, varying their service levels in order to get a good match between given resources and their computing requirements. This service level adaptation scheme is implemented for the streaming service on the camera and for some test services, performing mathematical operations. The resource manager is incorporated into systemd, and uses cgroups [16] to distribute the computing capacity. The experimental results show that the resource manager is fully operational and capable of managing and prioritizing resources as intended on the embedded system.