KubeGPU: efficient sharing and isolation mechanisms for GPU resource management in container cloud

Shen, Wenfeng; Liu, Zhengsen; Tan, Yunjie; Luo, Zhaokai; Zhou, Lei

doi:10.1007/s11227-022-04682-2

Cited by 4 publications

(2 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…KubeGPU also supports GPU resource isolation, ensuring that GPU resources between applications do not interfere with each other. The solution demonstrated good performance and scalability in experiments [39]. Chandra Prakash et al used dynamic slicing technology and virtual GPUs (vGPUs).Dynamic slicing technology are used to adjust the allocation ratio between CPU and GPU resources in real-time to achieve Goodput optimization.…”

Section: Gpu Remote-sharingmentioning

confidence: 99%

Evolution of GPU virtualization to resource pooling

liang,

Daud,

Ismail

2023

Second International Conference on Electronic Information Technology (EIT 2023)

View full text Add to dashboard Cite

With the rapid development of artificial intelligence, the demand for Graphics Processing Unit (GPU) resources is also increasing rapidly. To improve the efficiency and utilization of GPU resources, virtualization technology has been widely used in the field of GPUs. This article reviews the evolution from GPU virtualization to resource pooling, including device simulation, GPU pass-through, hardware-assisted virtualization, GPU full virtualization, GPU remotesharing, and GPU resource pools. The advantages and disadvantages of each stage and the challenges faced during the evolution process are also analyzed. This article discusses the difficulties and solutions of GPU resource pool construction, as well as the construction steps and framework, providing a reference for the research and application of GPU resource pooling.

show abstract

Section: Gpu Remote-sharingmentioning

confidence: 99%

Evolution of GPU virtualization to resource pooling

liang,

Daud,

Ismail

2023

Second International Conference on Electronic Information Technology (EIT 2023)

View full text Add to dashboard Cite

show abstract

“…The rapid evolution of cloud computing has significantly impacted the landscape of computational resource management, particularly with the integration of Graphics Processing Units (GPUs) in cluster environments [1,2,3]. GPUs provide substantial parallel processing capabilities, making them invaluable for Email addresses: siavashi@aut.ac.ir (Ahmad Siavashi), momtazpour@aut.ac.ir (Mahmoud Momtazpour) a wide range of computationally intensive applications [4].…”

Section: Introductionmentioning

confidence: 99%

GPU Cluster Dynamics: Insights from Alibaba's 2023 Trace Release

Siavashi,

Momtazpour

2023

Preprint

View full text Add to dashboard Cite

In this paper, we present a comprehensive analysis of GPU cluster traces from Alibaba, released in 2023, with a focus on unraveling the intricacies of node and pod configurations and their associated metrics. By dissecting the configurations of 1,523 nodes, predominantly GPU-based, we identify a wide array of configurations and GPU models, highlighting a balanced interplay between CPU cores and RAM across all nodes while also noting a decoupling between CPU/RAM and GPU-specific metrics in GPU nodes, indicating a nuanced approach to workload placement. Our investigation extends to an analysis of 8,152 pods, revealing diverse configurations and a significant prevalence of latency-sensitive pods, indicative of a cluster geared towards timely resource provisioning. The presence of Failed and Pending pods, however, underscores the complexity and challenges inherent in scheduling efficiency and resource allocation. By delving into both node and pod configurations and metrics, this study sheds light on the operational characteristics and workload patterns within the cluster. The insights gleaned from this analysis are invaluable for researchers, system designers, and operators, providing a nuanced understanding of the cluster's workings. This, in turn, lays the groundwork for future research and optimization efforts, ultimately contributing to the advancement of GPU cluster management and operation.

show abstract