Multi-core real-time scheduling for generalized parallel task models

Abstract

Abstract-Multi-core processors offer a significant performance increase over single core processors. Therefore, they have the potential to enable computation-intensive real-time applications with stringent timing constraints that cannot be met on traditional single-core processors. However, most results in traditional multiprocessor real-time scheduling are limited to sequential programming models and ignore intra-task parallelism. In this paper, we address the problem of scheduling periodic parallel tasks with implicit deadlines on multi-core processors. We first consider a synchronous task model where each task consists of segments, each segment having an arbitrary number of parallel threads that synchronize at the end of the segment. We propose a new task decomposition method that decomposes each parallel task into a set of sequential tasks. We prove that our task decomposition achieves a resource augmentation bound of 4 and 5 when the decomposed tasks are scheduled using global EDF and partitioned deadline monotonic scheduling, respectively. Finally, we extend our analysis to directed acyclic graph (DAG) task model where each node in the DAG has unit execution requirement. We show how these tasks can be converted into synchronous tasks such that the same transformation can be applied and the same augmentation bounds hold.

Patterns for Distributed Real-Time Stream Processing

“…Among the list of available approaches, some remarkable work are [17], [18], [19], [20], [35], which collectively deal with the needs of some of the most popular computing paradigms. In [17] the authors considered the multi-core scenario and proposed a parallel synchronous model inspired by the primitives of OpenMPI.…”

“…Another set of work ( [18], [19], [23], [60], [61], [62]) are related to the scheduling of parallel tasks on multi-core processors. Their computational models are mainly based on the fork-join tasks and the authors describe end-to-end interactions as a set of sequential and parallel steps.…”

An Execution Model for Fine-Grained Parallelism in Ada

“…We consider part of being a global scheduling approach that there is no priority inversion within the domain: at any given time, there is never a tasklet running on a processor in the dispatching domain if there are tasklets of tasks with a higher priority (or earlier deadline) awaiting execution. This model allows for the tasklet DAG to be converted ( Figure 6) to a DAG of subtasks ( Figure 6 middle) as commonly used in the real-time systems domain [16,17,18]. Response-time analysis techniques can also be used [19,20] (restricted to non-nested tasklets), as the tasklet model is actually more restricted than the general real-time DAG model, as it considers a fully-strict fork-join, thus a synchronous model can be used [18] (Figure 6 right).…”

Allocation of Parallel Real-Time Tasks in Distributed Multi-core Architectures Supported by an FTT-SE Network

“…The authors converted the parallel threads of a fork-join task into sequential tasks by creating a master thread, but with the difference (when compared to [5]) that no thread is ever allowed to migrate between cores. That work was generalized in [8], by allowing an arbitrary number of threads per parallel segment, and in [9] for the scheduling of tasks represented by a Directed Acyclic Graph (DAG).…”

“…Distributed systems have the particularity that the transmission delay of messages communicating threads within a task, cannot be deemed negligible as in the case of multi-core systems [5,7,8]. In here, we extend the problem of task allocation of fork-join real-time tasks presented in [1], by considering (i) a distributed multi-core architecture, and (ii) using a FTT-SE network for message transmission.…”