Minimizing CPU time shortage risks in integrated embedded software

Abstract

A major activity in many industries is to integrate software artifacts such that the functional and performance requirements are properly taken care of. In this paper, we focus on the problem of minimizing the risk of CPU time shortage in integrated embedded systems. In order to minimize this risk, we manipulate the start time (offset) of the software executables such that the system real-time constraints are satisfied, and further, the maximum CPU time usage is minimized. We develop a number of search-based optimization algorithms, specifically designed to work for large search spaces, to compute offsets for concurrent software executables with the objective of minimizing CPU usage. We evaluated and compared our algorithms by applying them to a large automotive software system. Our experience shows that our algorithms can automatically generate offsets such that the maximum CPU usage is very close to the known lower bound imposed by the domain constraints. Further, our approach finds limits on the maximum CPU usage lower than those found by a random strategy, and is not slower than a random strategy. Finally, our work achieves better results than the CPU usage minimization techniques devised by domain experts.

Identifying optimal trade-offs between CPU time usage and temporal constraints using search

“…Note that this value might not be a global optimal. However, based on our earlier experience [26], for the problem of minimizing CPU time usage in automotive applications, Hill Climbing performed best among several other singleobjective single-state algorithms. In addition, note that 1.43ms is a conservative lower bound, and we may never reach this value using our multi-objective search since the search in our case is constrained by the temporal coupling constraints.…”

“…Taking the former approach, we need to develop an efficient constraint model that can precisely specify a space of 10 320 solutions. Our earlier attempt to use constraint solvers resulted in a huge constraint model that did not fit in memory [26], and hence, the search could not even start (see Section 8 for more details). Note that we do not claim that our specific problem cannot be solved using constraint solvers.…”

“…In our earlier work [26], we provided formal conditions for checking satisfaction of the following restricted form of the temporal coupling constraint : The tasks in G have to run at the same time slot (i.e., k = 1) infinitely often. Using the chinese remainder theorem from classical number theory [30], we showed that the temporal coupling constraint for the case of k = 1 holds iff oi ≡ oj (modGCD(pi, pj)) for every ti, tj ∈ G (1) Using the same theorem, we compute at which time slots all of the tasks run together.…”

“…If this modification violates either SCS or the temporal coupling constraints, the tweak operator propagates the change to other offset values and modifies them appropriately so that in the end both constraints hold for the mutated Oj. In [26], we used a simpler version of this tweak operator (only for k = 1) in a single-objective single-state optimization algorithm. There, we showed that this tweak operator was more effective than our other choices.…”

“…Existing research focuses on real-time scheduling as a single-objective problem where the goal to be satisfied or optimized is to meet hard deadlines [7,13], or to optimize CPU time usage [27,26], or to check feasibility of real-time schedules [17,4,18], or to compute response time [10] and other timing properties [23]. We believe optimizing utilization of resources, such as optimizing CPU time usage, requires trade-off analysis and has to be formulated as a multi-objective problem.…”