POET: a portable approach to minimizing energy under soft real-time constraints

Abstract

Embedded real-time systems must meet timing constraints while minimizing energy consumption. To this end, many energy optimizations are introduced for specific platforms or specific applications. These solutions are not portable, however, and when the application or the platform change, these solutions must be redesigned. Portable techniques are hard to develop due to the varying tradeoffs experienced with different application/platform configurations. This paper addresses the problem of finding and exploiting general tradeoffs, using control theory and mathematical optimization to achieve energy minimization under soft real-time application constraints. The paper presents POET, an open-source C library and runtime system that takes a specification of the platform resources and optimizes the application execution. We test POET's ability to portably deliver predictable timing and energy reduction on two embedded systems with different tradeoff spaces -the first with a mobile Intel Haswell processor, and the second with an ARM big.LITTLE System on Chip. POET achieves the desired latency goals with small error while consuming, on average, only 1.3% more energy than the dynamic optimal oracle on the Haswell and 2.9% more on the ARM. We believe this open-source, librarybased approach to resource management will simplify the process of writing portable, energy-efficient code for embedded systems.

Automated control of multiple software goals using multiple actuators

“…Recent surveys capture the current state-of-the-art applying control-theory to software applications [17,46,58]-from controlling web server delays [38], to data service management [10], resource allocation [2,26,27,35], operating systems tuning [30,40,45], and energy management [25,41]. 1 http://www.martinamaggio.com/papers/fse17/ Some of these systems use automata-based formalisms to abstract software's behavior and temporal logic to specify some of its requirements [9,50], while we focus here on discrete-time control, where equation-based models are used to satisfy quantitative software properties.…”

A Probabilistic Graphical Model-based Approach for Minimizing Energy Under Performance Constraints

“…This duration gives sufficient time to measure system behavior. All applications are instrumented with the Application Heartbeats library which provides application specific performance feedback to LEO [22,27]. Thus LEO is ensured of optimizing the performance that matters to the application.…”

“…Given performance and power estimates, the energy minimization problem can be solved using existing convex optimization techniques [24,27,36,61]. LEO simply first take the estimates, then finds the set of configurations that represent Pareto-optimal performance and power tradeoffs, and (2) finally walks along the convex hull of this optimal tradeoff space until the performance goal is reached.…”

“…For example, many systems employ control theoretic designs which couple offline model building with online feedback control [9,24,27,36,39,46,47,51,57,61]. Over a narrow range of applications the combination of offline learning and control works well, as the offline models capture the general behavior of the entire class of application and require negligible online overhead.…”

Automated multi-objective control for self-adaptive software design

“…Several recent surveys capture the current state-of-the-art applying control-theory to software applications [5,33], as well as highlighting the main criticisms of early approaches [3]. Examples control delays for web servers [34], manage data centers [35], allocate resources [36][37][38], tune operating systems [39][40][41], minimize energy [42], and coordinate across the system stack [43]. These strategies adapt tunable knobs that can be identified either offline or at runtime [44].…”