Peak Power Management for scheduling real-time tasks on heterogeneous many-core systems

Abstract

The number and diversity of cores in on-chip systems is increasing rapidly. However, due to the Thermal Design Power (TDP) constraint, it is not possible to continuously operate all cores at the same time. Exceeding the TDP constraint may activate the Dynamic Thermal Management (DTM) to ensure thermal stability. Such hardware based closed-loop safeguards pose a big challenge in using many-core chips for real-time tasks. Managing the worst-case peak power usage of a chip can help toward resolving this issue. We present a scheme to minimize the peak power usage for frame-based and periodic real-time tasks on many-core processors by scheduling the sleep cycles for each active core and introduce the concept of a sufficient test for peak power consumption for task feasibility. We consider both inter-task and inter-core diversity in terms of power usage and present computationally efficient algorithms for peak power minimization for these cases, i.e., a special case of "homogeneous tasks on homogeneous cores" to the general case of "heterogeneous tasks on heterogeneous cores". We evaluate our solution through extensive simulations using the 48-core SCC platform and gem5 architecture simulator. Our simulation results show the efficacy of our scheme.

Peak-Power-Aware Primary-Backup Technique for Efficient Fault-Tolerance in Multicore Embedded Systems

“…In this paper, we assume that the system consists of k=m/2 core pairs CP={CP1, CP2, …, CPk}. The system executes a set of n non-preemptive frame-based soft real-time tasks Y={T1, T2, …, Tn} where the tasks release at the same time and share a common deadline D. Note, this assumption is valid for frame-based tasks and cases where multiple tasks belong to one complex multi-tasked application (e.g., video streaming), and such an application model is adopted in many related works such as [13] [30]. We also assume that each task Ti has an execution time τi and all tasks have the same period of D. Examples of these systems are medical care devices, avionics systems, control of chemical reactions, and surveillance systems [22], and an example of a frame-based soft real-time application is MPEG Player (multimedia) [31] [32].…”

“…However, replicated executions significantly increase system power consumption that may exceed the chip Thermal Design Power (TDP) constraint which cannot be sustained for longer periods of execution. TDP is considered as the highest sustainable power that a chip can dissipate before being forced to exploit a performance throttling mechanism, e.g., Dynamic Thermal Management (DTM) [13] [14]. DTM is a system-level solution to tackle thermal hot spots and to keep the system below a safe operating temperature.…”

“…DTM is a system-level solution to tackle thermal hot spots and to keep the system below a safe operating temperature. However, it may reduce system performance and may lead to violating system timing constraints [13] [14]. When a chip violates its TDP constraint, DTM techniques automatically trigger task migration or throttle the voltage and frequency to prevent hardware damage, but it can also significantly reduce the system's performance [13] [15].…”

“…However, it may reduce system performance and may lead to violating system timing constraints [13] [14]. When a chip violates its TDP constraint, DTM techniques automatically trigger task migration or throttle the voltage and frequency to prevent hardware damage, but it can also significantly reduce the system's performance [13] [15]. Therefore, DTM techniques, or at least their frequent triggers, may not be efficient for the systems that require satisfying strict timing constraints, e.g., real-time embedded systems [1] [13] [17] [18].…”

PkMin: Peak Power Minimization for Multi-Threaded Many-Core Applications

“…They propose an optimal algorithm that schedules a set of independent tasks from the application on a many-core to minimize application's peak power while meeting its deadline. The authors of [23] work on the same problem, but proposed an alternative light-weight heuristic algorithm. Most recently, the authors in [24] proposed an algorithm to minimize peak power for many-core applications with DAG under reliability constraints.…”