Ziad M. Kaakani

Integrated scheduling with garbage collection for real-time embedded applications in CLI

We present a schedulable garbage collection for realtime applications in virtual machine environments. The design objective is to make the pause time caused by garbage collection operations controllable, and the invocation of garbage collection predictable. Thus, real-time applications can be schedulable along with garbage collection. We develop a prototype for a schedulable garbage collection in MONO CLI execution environment. A cost model of garbage collection is established based on measured WCET to predict the execution time and overhead of garbage collection operations. A scheduling algorithm of garbage collection and application tasks is presented to illustrate how the time and memory constraints of real-time systems can be met. The experiment result of the scheduling algorithm for a periodic task set on the prototype is included in the paper

Schedulable persistence system for teal-time applications in virtual machine

Persistence in applications saves a computation state that can be used to facilitate system recovery upon failures. As we begin to adopt virtual execution environments (VMs) for mission-critical real-time embedded applications, persistence service will become an essential part of VM to ensure high availability of the systems.In this paper, we focus in a schedulable persistence system in VMs and show a prototype persistence system constructed on CLI s open source platform, MONO. By employing object serialization, the system enables concurrent and preemptible persistence operation, i.e., the task in charge of persistence service runs concurrently with application tasks and is a target of real-time scheduling. Thus, the execution of application tasks can be interleaved with the operations of persistence service, and the task timeliness can be guaranteed as the pause time caused by persistence service is bounded. The experiment output on the prototyped system illustrates that persistence service is appropriate for realtime applications because of its controllable pause time and its optimized overhead.