GENIE: QoS-guided Dynamic Scheduling for CNN-based Tasks on SME Clusters

Chen, Zhaoyun; Luo, Lei; Yang, Haoduo; Yu, Jie; Wen, Mei; Zhang, Chunyuan

doi:10.23919/date.2019.8715279

Cited by 7 publications

(3 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…• GENIE: A prediction-based heuristic scheduler. 11 The prediction model according to the domain-specific information of DL tasks is derived from a comprehensive characterization. GENIE dynamically identifies the best placements for DL tasks and schedules them on the GPU cluster based on the shortest waiting allowance first (SWAF) principle.…”

Section: Baselinementioning

confidence: 99%

“…4,6 For example, research 4 explores similar patterns of diverse tasks based on historical information and batching tasks in order to improve system efficiency. Moreover, researches 11,12 construct a performance prediction model based on DL workload characterization to guide their heuristic scheduling algorithms. While these prediction-based schedulers can improve system efficiency, they also face some challenges in DL R&D scenarios, including the following:…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

RIFLING: A reinforcement learning‐based GPU scheduler for deep learning research and development platforms

Chen

2021

Softw Pract Exp

Self Cite

View full text Add to dashboard Cite

GPU platforms have been widely adopted in both academia and industry to support deep learning (DL) research and development (R&D). Compared with giant companies who favor custom-designed AI platforms, most small-and-medium-sized enterprises, institutes and universities (EIUs) prefer to build or rent a cost-effective GPU cluster, usually in a limited-scale, to process diverse DL R&D workloads. Therefore, more attention has been attracted by DL scheduling with the aim of improving the system efficiency and task performance. However, prior prediction-based schedulers are limited in terms of their prediction accuracy and profiling overhead. Accordingly, in this article, we propose a reinforcement learning (RL)-based online GPU scheduler, RIFLING, to model the scheduling problem as an online decision-making process. Scheduling decisions are made according to Q-learning, which is a typical RL method. RIFLING can achieve high scheduling efficiency based on the online exploring and exploiting of diverse scheduling strategies for various DL workloads, without the need for expensive offline profiling or sophisticated prediction model. We implement RIFLING as a plugin of Tensorflow, and deploy it on a distributed GPU cluster. Experiments demonstrate that RIFLING achieves up to 47.8% reductions and 19.6% improvements in makespan and average normalized processing rate respectively compared to the best available baseline without any manual intervention.

show abstract

Section: Baselinementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

RIFLING: A reinforcement learning‐based GPU scheduler for deep learning research and development platforms

Chen

2021

Softw Pract Exp

Self Cite

View full text Add to dashboard Cite

show abstract

“…This inescapability is applied to forecast whether a job can be talented before its target also guesses suitable resource formation. However, this approach can't satisfy the QoS services [Chen et al, 2019]. Multi Dimension-Affinity Propagation Clustering approach is applying the ML to decide the graininess.…”

Section: Related Workmentioning

confidence: 99%

Machine Learning-Based Efficient Clustering and Improve Quality of Service in Manet

Reddy¹,

Mungara²

2021

INDJCSE

View full text Add to dashboard Cite

A Mobile Ad hoc Network (MANET) is a self-governing system in that numerous nodes are connected with each other, applying multi-hop wireless links. MANET has overturned society owing to their infrastructure-less decentralized modes of transaction. As a result, researchers have concentrated on discovering better directions to operate the probability of MANETs completely. The current initiation of machine learning methods has prepared it to concern artificial intelligence to increase the best approaches for this function. In this work, Machine Learning-based Efficient Clustering and Improve Quality of Service in MANET (MLEC) is proposed. In this work, we improve the clustering technique that applied Quality of Service (QoS) parameters to choose Cluster Heads (CHs). The cluster formation and CH selection is accessible via a multi-objective fitness function applying Particle Swarm Optimization (PSO) algorithm. This approach gives three inputs: node mobility, energy utilization rate, and bandwidth. The algorithm establishes a higher rate of achievement in forecasting output values also could correctly recognize better CH that should offer the well-organized routing. In addition, it reduces the intra-cluster distance among nodes and their relevant CHs. The proposed work is experimented with broadly in the network simulator and compared with the other baseline protocols.

show abstract