RL-Based Scheduling Strategies in Actual Grid Environments

Costa, Bruno; Dutra, Inês; Mattoso, Marta

doi:10.1109/ispa.2008.119

Cited by 3 publications

(5 citation statements)

References 13 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The scheduling strategy presented in References 38, 39 inspired the reward function used in the current work. However, while there they aimed at job orchestration in computational grids, here we focus on finding a plan for scheduling scientific workflows activities executed in a cloud environment.…”

Section: Related Workmentioning

confidence: 99%

“…The only requirement is that the agent should be able to evaluate if the current schedule is “better” or “worse” than a previous one. There are some approaches that apply RL to scheduling 38‐40 but they are decoupled from the concept of scientific workflows (although some of them model an application as a directed acyclic graph).…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

An incremental reinforcement learning scheduling strategy for data‐intensive scientific workflows in the cloud

Nascimento

Silva

Paes

et al. 2021

Concurrency and Computation

View full text Add to dashboard Cite

Summary Most scientific experiments can be modeled as workflows. These workflows are usually computing‐ and data‐intensive, demanding the use of high‐performance computing environments such as clusters, grids, and clouds. This latter offers the advantage of the elasticity, which allows for changing the number of virtual machines (VMs) on demand. Workflows are typically managed using scientific workflow management systems (SWfMS). Many existing SWfMSs offer support for cloud‐based execution. Each SWfMS has its scheduler that follows a well‐defined cost function. However, such cost functions should consider the characteristics of a dynamic environment, such as live migrations or performance fluctuations, which are far from trivial to model. This article proposes a novel scheduling strategy, named ReASSIgN, based on reinforcement learning (RL). By relying on an RL technique, one may assume that there is an optimal (or suboptimal) solution for the scheduling problem, and aims at learning the best scheduling based on previous executions in the absence of a mathematical model of the environment. For this, an extension of a well‐known workflow simulator WorkflowSim is proposed to implement an RL strategy for scheduling workflows. Once the scheduling plan is generated via simulation, the workflow is executed in the cloud using SciCumulus SWfMS. We conducted a throughout evaluation of the proposed scheduling strategy using a real astronomy workflow named Montage.

show abstract

Section: Related Workmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

An incremental reinforcement learning scheduling strategy for data‐intensive scientific workflows in the cloud

Nascimento

Silva

Paes

et al. 2021

Concurrency and Computation

View full text Add to dashboard Cite

show abstract

“…The corresponding software, hardware or combined faults can cause resource unavailability [16]. In [23] two scheduling strategies for grid environments based on reinforcement learning are reviewed which are named modified MQD and AG. Then these two strategies are compared with a standard scheduling algorithm based on round-robin (RR).…”

Section: Related Workmentioning

confidence: 99%

“…In a first step, subtasks are ordered according to their estimated execution time. Shortest subtasks are selected first, and, in [23] implementation, assigned to resources in a roundrobin fashion. This step terminates when all queues have their limit exceeded.…”

Section: Related Workmentioning

confidence: 99%

“…In this simulation we supposed, number of necessary entities as follow: [3][4][5][6][7][8] Figure 3 shows make span comparison of modified MQD (Multiple Queue with Duplication) and LESG. As [23] we compute make span as equation (11). ttr tf te span make Eq(11) In this equation te is task execution time at the remote place, tf is queue waiting time and ttr is time to transfer input and output.…”

Section: Simulationmentioning

confidence: 99%

See 1 more Smart Citation

LESG: Learning and economic based scheduler implementation

Khanli

Fard

2009

2009 2nd IEEE International Conference on Broadband Network &Amp; Multimedia Technology

View full text Add to dashboard Cite

In a dynamic environment like grid with huge number of subtasks, flexible approach is necessary to manage resource allocation.Grid is a robust technique in parallel computing. The central component of grid is resource management system (RMS). The main functions of RMS are scheduling and allocation of subtasks. The goal of this paper is to provide an optimal learning solution for dynamically choosing appropriate resource. In this paper we introduce an intelligent approach to schedule subtasks based on reinforcement learning. That is named LESG. In LESG a flexible allocation according to subtasks and resources attributes, increases performance of gird.

show abstract