On the scheduling of fork-join parallel/distributed real-time tasks

Abstract

Modern real-time embedded applications present high computation requirements which need to be realized within strict time constraints. The current trend towards parallel processing in the embedded domain allows providing higher processing power. However, in some embedded applications, the use of powerful enough multi-core processors, may not be possible due to energy, space or cost constraints. A solution for this problem is to extend the parallel execution of the applications, allowing them to distribute their workload among networked nodes, on peak situations, to remote neighbour nodes in the system. In this context, we present the Partitioned-Distributed-Deadline Monotonic Scheduling algorithm for fork-join parallel/distributed fixed-priority tasks. We study the problem of scheduling fork-join tasks that execute in a distributed system, where the inherent transmission delay of tasks must be considered and cannot be deemed negligible, as in the case of multicore systems. Our scheduling algorithm is shown to have a resource augmentation bound of 4, which implies that any task set that is feasible on m unit-speed processors and a single shared real-time network, can be scheduled by our algorithm on m processors and a single real-time network that are 4 times faster. We confirm through simulations our analytical results.

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

“…In the previous work [1] we studied the problem of scheduling fork-join tasks on a distributed system composed of single-processor nodes and a shared bus communication network. 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].…”

“…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.…”

“…In a similar manner as in [1], we transform threads composing a P/D task into a set of independent sequential tasks with constrained deadlines. This transformation is based on the imposition of a set of artificial intermediate deadlines (denoted as d i,j ), to threads θ i,j,k and messages µ i,j,k , in each segment σ i,j .…”

“…The fork-join Parallel/Distributed real-time model (P/D tasks) [1], was designed to consider such execution pattern. In this paper, we consider P/D tasks and a distributed computing platform composed of multi-core nodes, and interconnected by a Flexible Time Triggered -Switched Ethernet (FTT-SE) network [2].…”

Holistic analysis for fork-join distributed tasks supported by the FTT-SE protocol

“…Future applications will require a network that conciliates real-time traffic guarantees (time-triggered and event-triggered), best-effort traffic, on-line scheduling, etc., and also they will require the use of more powerful computing models. For those reasons, in this work we propose the use of a Flexible Time Triggered -Switched Ethernet network (FTT-SE) [2], and the use of distributed real-time applications which are composed of a set of fork-join Parallel/Distributed real-time tasks (P/D tasks) [3].…”