Citation Link: https://nbn-resolving.org/urn:nbn:de:hbz:467-14715
Scenario-based meta-scheduling for energy-efficient, robust and adaptive time-triggered multi-core architectures
Source Type
Doctoral Thesis
Author
Issue Date
2019
Abstract
Complex electronic systems are used in many safety-critical applications (e.g., aerospace, automotive, nuclear stations), for which certification standards demand the use of assured design methods and tools. Scenario-based meta-scheduling (SBMeS) is a way of managing the complexity of adaptive systems via predictable behavioral patterns established by static scheduling algorithms. SBMeS is highly relevant to the internet of things (IoT) and real-time systems. Real-time systems are often based on time-triggered operating systems and networks and can benefit from SBMeS for improved energy-efficiency, flexibility and dependability.
This thesis introduces an SBMeS algorithm that computes an individual schedule for each relevant combination of events such as dynamic slack occurrences. Dynamic frequency scaling of cores and routers is used to improve energy efficiency while preserving the temporal correctness of time-triggered computation and communication activities (e.g., collision avoidance, timeliness). Models of applications, platforms and context are used by scheduling tools to prepare reactions to events and to generate meta-schedules.
In this work, techniques and tools are developed to schedule a set of tasks and messages on Network-on-chip (NoC) architectures to minimize total energy consumption, considering time constraints and adjustable frequencies. This algorithm is intended for mixed-criticality and safety-critical adaptive time-triggered systems and can cover fault-tolerance requirements. It can also help to react to fault events by recovering the system. We also introduce a meta-scheduling visualization tool (MeSViz) for visualizing schedules. We also introduce a meta-scheduling visualization tool (MeSViz) for visualizing schedules.
We experimentally and analytically evaluate the schedules’ energy-efficiency for cores and routers. In addition, the timing is analytically evaluated, based on static slack and dynamic slack events. Simulation results show that our dynamic slack algorithm produces, on average, an energy reduction of 64.4% in a single schedule and 41.61% energy reduction for NoCs. By compressing the schedule graphs the memory consumption can be reduced by more than 61%.
This thesis introduces an SBMeS algorithm that computes an individual schedule for each relevant combination of events such as dynamic slack occurrences. Dynamic frequency scaling of cores and routers is used to improve energy efficiency while preserving the temporal correctness of time-triggered computation and communication activities (e.g., collision avoidance, timeliness). Models of applications, platforms and context are used by scheduling tools to prepare reactions to events and to generate meta-schedules.
In this work, techniques and tools are developed to schedule a set of tasks and messages on Network-on-chip (NoC) architectures to minimize total energy consumption, considering time constraints and adjustable frequencies. This algorithm is intended for mixed-criticality and safety-critical adaptive time-triggered systems and can cover fault-tolerance requirements. It can also help to react to fault events by recovering the system. We also introduce a meta-scheduling visualization tool (MeSViz) for visualizing schedules. We also introduce a meta-scheduling visualization tool (MeSViz) for visualizing schedules.
We experimentally and analytically evaluate the schedules’ energy-efficiency for cores and routers. In addition, the timing is analytically evaluated, based on static slack and dynamic slack events. Simulation results show that our dynamic slack algorithm produces, on average, an energy reduction of 64.4% in a single schedule and 41.61% energy reduction for NoCs. By compressing the schedule graphs the memory consumption can be reduced by more than 61%.
File(s)![Thumbnail Image]()
Loading...
Name
Dissertation_Babak_Sorkhpour.pdf
Size
6.52 MB
Format
Adobe PDF
Checksum
(MD5):b4ca88d48835cba8ef8289e3aabff35b
Owning collection