Integer and combinatorial optimization
Integer and combinatorial optimization
Scheduling with optimized communication for time-triggered embedded systems
CODES '99 Proceedings of the seventh international workshop on Hardware/software codesign
Scheduling of Time-Triggered Real-Time Systems
Constraints
Advanced model transformation language constructs in the VIATRA2 framework
Proceedings of the 2006 ACM symposium on Applied computing
Model-based optimization of enterprise application and service deployment
ISAS'05 Proceedings of the Second international conference on Service Availability
| Hi-index | 0.00 |
Distributed system composition is the main trend in creating safety-critical (SC) real-time systems like automotive, aerospace, and industrial control systems. Their growing complexity (e.g. tens of control units in a modern car) led to an integrated architecture concept [7]. It supports the sharing of hardware resources between different sub-applications for the sake of cost reduction, but still keeps the overall system safety by properly isolating jobs from each other. Validation and certification of SC sytems are a key problem. They are especially hard, if not impossible at all, if the behavior of the system is non-deterministic. The time triggered (TT) paradigm (such as TTP/C [14] and FlexRay [4]) uses a strictly deterministic, static, design time generated schedule for both the computation jobs in the processing nodes and the internode communication tasks. Current tools create the intranode job and interjob communication allocation and scheduling in two distinct steps in order to reduce the total computational complexity to a feasible level. However, this separation of the two design steps despite their strong mutual influence may result in sub-optimal resource utilization, thus additional costs. The rapid growth in the computational power commonly available to the designer justifies re-visiting the potential of single phased global optimization. The recent paper introduces a novel approach calculating resource allocation and task schedules in a single step by using a standard mixed integer linear programming (MILP) solver covering extra-functional requirements as well. At first, optimization is used to explore the boundaries of the design space from the points of view of cost, throughput, robustness and extensibility. Subsequently, the designer can formulate his priorities between these, frequently contradicting goals by creating a weighted objective funtion. Optimization is accelerated by heuristic lower and upper estimates.