Optimizing Resource Usage in Component-Based Real-Time Systems

Abstract

Abstract. The embedded systems domain represents a class of systems that have high requirements on cost efficiency as well as run-time properties such as timeliness and dependability. The research on componentbased systems has produced component technologies for guaranteeing real-time properties. However, the issue of saving resources by allocating several components to real-time tasks has gained little focus. Trade-offs when allocating components to tasks are, e.g., CPU-overhead, footprint and integrity. In this paper we present a general approach for allocating components to real-time tasks, while utilizing existing real-time analysis to ensure a feasible allocation. We demonstrate that CPU-overhead and memory consumption can be reduced by as much as 48% and 32% respectively for industrially representative systems.

Architecture optimization: speed or accuracy? both!

“…A relevant body of these solutions aim at only satisfying predefined constraints (Martens and Koziolek 2009), while the rest seeks a near optimal deployment candidate without violating a set of given constraints (Medvidovic and Malek 2007;Fredriksson et al 2005). Model-driven approaches usually provide good approximations, which are although often not enough for embedded systems where runtime measurements are needed in order to assess certain critical performance-related quality attributes.…”

Solving a real-time allocation problem with constraint programming

“…For models quite similar to the one presented in Section 2, [63,11,41,54,4,50,21] are good examples of the research efforts. Like us, they all consider a fixed priority scheduler.…”

“…Allocation and scheduling are usually considered as two independent stages. For most of them [63,11,41,54,4,21], only the allocation problem is solved since task priorities are known. In [50], Richard et al propose a method that simultaneously allocates tasks to processors and assigns priorities to tasks and messages.…”

“…Different techniques are used: a simulated annealing algorithm in [63,11], a branch-and-bound search algorithm in [41,50], a genetic algorithm in [21], and a dedicated heuristic in [54]. In [4], a clustering algorithm is proposed.…”

“…Since the problem of allocating tasks is generally NP-hard [34], some form of enumerative method or approximation using heuristics needs to be developed for this problem: graph theory techniques [57,38,18], branch-andbound [41,45,44,50,65], genetic algorithm [40,6,53,2,39,19,20,42,21], clustering [4,1,36], steepest descent (or Hill climbing) [40,64], tabu search [46,64]; simulated annealing [63,11,40,14,15,17,19,64], neural network [56,2], and dedicated heuristics [54,17,32,66,3]. However, today, for a specific problem, no technique seems to be more appropriate than another.…”