Online Scheduling of Task Graphs on Hybrid Platforms

Abstract

Abstract. Modern computing platforms commonly include accelerators. We target the problem of scheduling applications modeled as task graphs on hybrid platforms made of two types of resources, such as CPUs and GPUs. We consider that task graphs are uncovered dynamically, and that the scheduler has information only on the available tasks, i.e., tasks whose predecessors have all been completed. Each task can be processed by either a CPU or a GPU, and the corresponding processing times are known. Our study extends a previous 4 m/k-competitive online algorithm [2], where m is the number of CPUs and k the number of GPUs (m ≥ k). We prove that no online algorithm can have a competitive ratio smaller than m/k. We also study how adding flexibility on task processing, such as task migration or spoliation, or increasing the knowledge of the scheduler by providing it with information on the task graph, influences the lower bound. We provide a (2 m/k + 1)-competitive algorithm as well as a tunable combination of a system-oriented heuristic and a competitive algorithm; this combination performs well in practice and has a competitive ratio in Θ( m/k). Finally, simulations on different sets of task graphs illustrate how the instance properties impact the performance of the studied algorithms and show that our proposed tunable algorithm performs the best among the online algorithms in almost all cases and has even performance close to an offline algorithm.

A State-of-the-Art Review of Task Scheduling for Edge Computing: A Delay-Sensitive Application Perspective

“…The foundation of a task scheduler algorithm for real-time services is to assemble an appropriate graph of tasks, a directed acyclic graph (DAG), consisting of serial and parallel tasks. However, scheduling tasks based on the DAG representing a real-time application is an NP-hard problem [50].…”

Online Scheduling of Task Graphs on Heterogeneous Platforms

“…• We study the generalized problem with multiple types of processors, on which adapt the lower bounds and the online algorithm (Section 6); • We propose a simple heuristic (Section 4.2) based on QA and the system-oriented heuristic EFT, which is both a competitive algorithm and performs well in practice, as we show with a comprehensive simulation set (Section 7). Note that a preliminary version of this work was presented at the Euro-Par 2018 conference [2].…”

Analysis of a List Scheduling Algorithm for Task Graphs on Two Types of Resources

“…Amaris et al proposed an online 4 max( m n , n m )-approximation algorithm based on comparison of processing times on both resources [21], which is similar to [14], [15] approaches for independent tasks. This algorithm and its analysis have been recently improved to obtain a 1 + 2 max( m n , n m ) ratio [3].…”

“…We perform comparison of HeteroPrioDep with HLP [21], ER-LS [24] , QA [3], HEFT [19] and ECT [3] algorithms from the literature. ER-LS, QA and ECT are online algorithms while HLP and HEFT are offline algorithms.…”

“…Third, scheduling is a well known NP-complete optimization problem and hybrid resources make this problem harder. We point the reader to [2] for a study on the complexity of scheduling problems and to [3] for a recent study in the case of hybrid platforms. All these observations led to the development of different task based runtime systems.…”