Dynamic load shedding mechanisms for overloads in distributed embedded systems

Abstract

The authors present dynamic load shedding mechanisms that support multi-grain parallelism in distributed embedded systems. An approach is presented to provide overload detection and load reconfiguration mechanisms within concurrent languages. The approach makes it possible to have adaptive resource control constructs dynamically reconfigure application parameters based on the overload condition. This capability allows for more complex resource control policies that utilize information about the actual application being performed. A set of resource control components is provided within a language runtime environment that supports load balancing in multi-grain distributed embedded systems. Overload resource requirements can be managed effectively within the concurrent language runtime environment.<>