The scheduling based on machine learning for heterogeneous CPU/GPU systems

Shulga, D. A.; Kapustin, A. A.; Kozlov, A.; Kozyrev, A. A.; Rovnyagin, Mikhail M.

doi:10.1109/eiconrusnw.2016.7448189

Cited by 22 publications

(11 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Their main focus is to achieve power-aware scheduling on CPU-GPU environments. On the other hand, Shulga et al [40] propose a scheduler that selects targets to execute according to machine learning-based training with different data sizes. In [30], authors propose visual analysis techniques to evaluate the execution time of high-performance applications on hybrid architectures.…”

Section: Related Workmentioning

confidence: 99%

Scheduling for heterogeneous systems in accelerator-rich environments

Yesil

Öztürk

2021

J Supercomput

View full text Add to dashboard Cite

The world is creating ever more data and the applications are required to deal with ever-increasing datasets. To process such datasets heterogeneous and manycore accelerators are being deployed in various computing systems to improve energy efficiency. In this work, we present a runtime management system designed for such heterogeneous systems with manycore accelerators. More specifically, we design a resource-based runtime management system that considers application characteristics and respective execution properties on the nodes and accelerators. We propose scheduling heuristics and run time environment solutions to achieve better throughput and reduced energy in computing systems with different accelerators. We give implementation details about our framework; show different scheduling algorithms, and present experimental evaluation of our system. We also compare our approaches with an optimal scheme where integer linear programming approach has been implemented for mapping applications on the heterogeneous system. While it is possible to extend the proposed framework to a wide variety of accelerators, our initial focus is on Graphics Processing Units (GPUs). Our experimental evaluations show that including accelerator support in the management framework improves energy consumption and execution time significantly. We believe that this approach has the potential to provide an effective solution for next generation accelerator-based computing systems.

show abstract

Section: Related Workmentioning

confidence: 99%

Scheduling for heterogeneous systems in accelerator-rich environments

Yesil

Öztürk

2021

J Supercomput

View full text Add to dashboard Cite

show abstract

“…Learning is performed via back-propagation that adjusts the weights and finds an optimum which minimizes the prediction error based on observed target output and predicted variable. Due to their lightweight implementation and high accuracy, in the past, ANNs were used in core performance prediction and thread partitioning strategies [21], [27], [35], [36]. Updates to ANN require adjusting the model weights while updates to OLS and EN models necessitate recomputation of regression coefficients.…”

Section: Training Of Per-replica Performance Prediction Modelsmentioning

confidence: 99%

SEER: Performance-Aware Leader Election in Single-Leader Consensus

Sakic,

Vizarreta,

Kellerer

2021

Preprint

View full text Add to dashboard Cite

Modern stateful web services and distributed SDN controllers rely on log replication to omit data loss in case of fail-stop failures. In single-leader execution, the leader replica is responsible for ordering log updates and the initiation of distributed commits, in order to guarantee log consistency. Network congestions, resource-heavy computation, and imbalanced resource allocations may, however, result in inappropriate leader selection and increased cluster response times.We present SEER, a logically centralized approach to performance prediction and efficient leader election in leader-based consensus systems. SEER autonomously identifies the replica that minimizes the average cluster response time, using prediction models trained dynamically at runtime. To balance the exploration and exploitation, SEER explores replicas' performance and updates their prediction models only after detecting significant system changes. We evaluate SEER in a traffic management scenario comprising [3..7] Raft replicas, and well-known datacenter and WAN topologies. Compared to the Raft's uniform leader election, SEER decreases the mean control plane response time by up to ∼ 32%. The benefit comes at the expense of the minimal adaptation of Raft election procedure and a slight increase in leader reconfiguration frequency, the latter being tunable with a guaranteed upper bound. No safety properties of Raft are invalidated by SEER.

show abstract

“…Furthermore, Wen et al 17 introduced a scheme for OPENCL scheduler that can schedule multiple programs on a heterogeneous computing system using a prediction technique. Schulga et al 18 based the design of the scheduler in a machine-learning algorithm.…”

Section: Related Workmentioning

confidence: 99%

KubCG: A dynamic Kubernetes scheduler for heterogeneous clusters

2020

View full text Add to dashboard Cite

SummaryContainer platforms are increasingly being used to deploy cloud‐based services. Nevertheless, many cloud services are also demanding graphics processing units (GPUs) to accelerate different applications that make use of their parallel architecture, such as deep learning or just video processing. Thus, different container technologies, such as Docker and Kubernetes, are implementing GPU support. Some effort is being devoted to design algorithms to schedule applications into heterogeneous computing systems that use CPUs and GPUs together. This article is part of this effort, and we describe how to build a dynamic scheduling platform for Kubernetes that is able to manage the deployment of Docker containers in a heterogeneous cluster, which we call KubCG. This platform implements a new scheduler that optimizes the deployment of new containers by taking into account the Kubernetes Pod timeline and the historical information about the execution of the containers. We have performed different tests to validate this new algorithm, and KubCG was able to reduce the time to complete different tasks down to a 64% of the original time in our different experiments.

show abstract

The scheduling based on machine learning for heterogeneous CPU/GPU systems

Cited by 22 publications

References 6 publications

Scheduling for heterogeneous systems in accelerator-rich environments

Scheduling for heterogeneous systems in accelerator-rich environments

SEER: Performance-Aware Leader Election in Single-Leader Consensus

KubCG: A dynamic Kubernetes scheduler for heterogeneous clusters

Contact Info

Product

Resources

About