Adaptive Framework for Efficient Resource Management in RTOS

Embedded systems, applications and the environment that they are deployed in have all become increasingly complex in recent years. Application demands for more resources and dynamic changes in the environment make resource management in Real-Time Operating Systems (RTOS) extremely challenging. Key approaches include specialisation or adaptation of the RTOSs resource management policies according to the dynamic requirements of the applications. This paper describes a reflection-based adaptive RTOS framework that allows dynamic application driven adaptation of RTOS resource management policies. The context assumed by this paper is that of low-cost, limited-resource embedded systems with modern hardware i.e. complex CPUs with support for virtual memory. Also, when developing a general purpose RTOS, there is limited knowledge of the potential applications that will use the RTOS. Thus, the RTOS is built for the general-case rather than according to application-specific requirements. Such an RTOS implements generic resource management policies. This paper discusses the issues related to providing application-specific resource management in a general purpose RTOS. An Existing approach of defining an adaptive framework in the RTOS using reflection is described along with some results of its implementation in Linux (2.6.16 kernel). The paper also discusses the possible future directions to the approach.