Browsing by Subject "Regularity-based resource partitioning"

Now showing 1 - 1 of 1

Resource scheduling and design of real-time Cyber-Physical Systems in the open system environment
(2020-06-25) Chen, Wei-Ju; Mok, Aloysius Ka-Lau; Rossbach, Christopher J.; Qiu, Lili; Han, Song
In this thesis, we consider the problem of resource provisioning and system designs for real-time Cyber-Physical Systems (CPS) in an open system environment. In an open system environment, there does not exist a global resource scheduler that has complete knowledge of the real-time performance requirements of each individual application that shares the resources with the other applications. Regularity-based Resource Partition (RRP) model is an effective strategy to hierarchically partition and allocate resource to each application. Although RRP model is proved to be effective for such system, there is still a gap between the real system design and the model. Extant RRP model only discusses uniform resource environment where the size of each unit of resource supply (a resource slice) is the same. In a real world application, resource slices from different physical resources may have different size. This may incur unexpected delay and deadline misses for tasks. Moreover, extant RRP model does not consider changes in resource requests from the applications at run time. How the real system can be designed to utilize RRP model in such dynamic settings is also not well-studied. We aim to address these issues by the followings. (1) We extend the RRP model to non-uniform multi-resource environments by introducing a novel composite resource partition abstraction, identifying the feasible scheduling conditions for regular composite resource partitioning and propose an Acyclic Regular Composite Resource Partition Scheduling (ARCRPS) algorithm. (2) We formalize the Dynamic Partition Reconfiguration (DPR) problem for resource reconfiguration by introducing the concept of reconfiguration regularity as the performance semantics during the reconfiguration and propose a novel 3-stage algorithm to solve the DPR problem. (3) Based on the above mentioned model extension, we explore the two problems by applying the model to real systems. We demonstrate the proposed work on a networked system and a real autonomous F1/10 model car system to show the applicability of the model.