@misc{14037,
  author = {Stefano Di Alesio},
  title = {Optimal Performance Tuning in Real-Time Systems using Multi-objective Constrained Optimization},
  abstract = {Real-Time Embedded Systems (RTES) in safety-critical applications have to meet strict performance requirements to be deemed safe for operation. The satisfaction of these requirements at runtime often depends on configuration parameters that regulate how software tasks interact with hardware sensors and actuators. Tuning performance-related parameters is usually a manual, time-consuming, and error-prone process. This is because these parameters and their values define a large space of system configurations, and evaluating how each configuration affects the performance often requires executing the whole system. In this paper, we express RTES performance tuning as a multi-objective Constrained Optimization Problem (COP) over the configuration space that captures the dependencies between configuration parameters and performance requirements. In this way, the COP solutions characterize configurations predicted to maximize the satisfaction of performance requirements, and can in turn be used as guidelines for optimal performance tuning. We develop the COP as an OPL model for IBM ILOG CP OPTIMIZER, and validate our approach on a safety-critical I/O drivers system from the maritime and energy domain. The validation shows that our approach identifies within half an hour configurations characterized by tasks delay times that minimize deadline misses, response time, and CPU usage.},
  year = {2016},
  journal = {The 22nd International Conference on Principles and Practice of Constraint Programming (CP 2016)},
  doi = {10.1007/978-3-319-44953-1_35},
}