Interference-Aware Orchestration in Kubernetes

Tzenetopoulos, Achilleas; Masouros, Dimosthenis; Xydis, Sotirios; Soudris, Dimitrios

doi:10.1007/978-3-030-59851-8_21

Cited by 14 publications

(6 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The heterogeneous Hardware infrastructure is integrated into a single execution environment using Kubernetes [40], [41] with The Skynet [42], Volcano [43], and the interferenceaware custom scheduler [44] extensions. For the unified storage layer EVOLVE uses open-source software frameworks such as Karvdash [45], Datashim [46], and H3 [47].…”

Section: Big Data and High-performance Computing Platform Evolvementioning

confidence: 99%

Object Detection in Driving Datasets Using a High-Performance Computing Platform: A Benchmark Study

et al. 2022

View full text Add to dashboard Cite

Nowadays, machine learning methods are increasingly used in different parts of autonomous driving and driving assistance systems. Yet, data and computational requirements can be enormous with these methods. Thus, providing several datasets containing many and diverse cases for the target problem and sufficient hardware for training and application of ML methods are too critical for achieving accurate results when applying them. Hence, we present an object detection benchmark study implementing the knowledge graph-based data integration framework to meet the data requirements and run the implementation on a big data and high-performance computing (HPC) platform, namely the EVOLVE. We applied different object detection methods to widely known open datasets, and compared the results on three different hardware setups, including EVOLVE. We also performed a small-scale transfer learning experiment. The results show that EVOLVE allowed the exploitation of much bigger data leading to a more efficient application of the object detection models with the help of the knowledge graph-based data integration framework. EVOLVE significantly improved the execution times compared to running them on a local laptop and a virtual machine and provided the easy-to-use and ready-to-use means to store large datasets and apply different models with its hardware and software stack.

show abstract

Section: Big Data and High-performance Computing Platform Evolvementioning

confidence: 99%

Object Detection in Driving Datasets Using a High-Performance Computing Platform: A Benchmark Study

et al. 2022

View full text Add to dashboard Cite

show abstract

“…The rise of "cloud-native" platforms, such as Kubernetes, that facilitate the deployment of applications on lightweight containers and expand their capacity to dynamically scale resources, have provided grounds for the study of their impact on the performance of applications. 8,12 A portion of prior scientific works propose novel Kubernetes schedulers to optimize the placement of incoming containerized applications on the underlying cluster, [20][21][22][23] either by minimizing the overall energy utilization of cloud/edge nodes, 20 or by attempting to reduce the interference between co-located workloads either on the system or the network level. [21][22][23] Moreover, except for Kubernetes-based resource management approaches, other scientific papers propose custom solutions able to efficiently place workloads on edge nodes.…”

Section: Related Workmentioning

confidence: 99%

“…8,12 A portion of prior scientific works propose novel Kubernetes schedulers to optimize the placement of incoming containerized applications on the underlying cluster, [20][21][22][23] either by minimizing the overall energy utilization of cloud/edge nodes, 20 or by attempting to reduce the interference between co-located workloads either on the system or the network level. [21][22][23] Moreover, except for Kubernetes-based resource management approaches, other scientific papers propose custom solutions able to efficiently place workloads on edge nodes. In Amit Samanta and Jianhua Tang's "Dyme," the authors propose a dynamic microservice scheduling scheme for mobile edge computing, aiming to reduce energy consumption while providing fair Quality-of-Service (QoS) among all devices.…”

Section: Related Workmentioning

confidence: 99%

HLF-Kubed: Blockchain-Based Resource Monitoring for Edge Clusters

Tzenetopoulos

Masouros

Kapsoulis

et al. 2022

ledger

View full text Add to dashboard Cite

In the past several years, there has been an increased usage of smart, always- connected devices at the edge of the network, which provide real-time contextual information with low overhead to optimize processes and improve how companies and individuals interact, work, and live. The efficient management of this huge pool of devices requires runtime moni- toring to identify potential performance bottlenecks and physical defects. Typical solutions, where monitoring data are aggregated in a centralized manner, soon become inefficient, as they are unable to handle the increased load and become single points of failure. In addition, the resource-constrained nature of edge devices calls for low-overhead monitoring systems. In this paper, we propose HLF-Kubed, a blockchain-based, highly available framework for monitoring edge devices, leveraging distributed ledger technology. HLF-Kubed builds upon Kubernetes container orchestrator and HyperLedger Fabric frameworks and implements a smart contract through an external chaincode for resource usage storing and querying. Our experimental results show that our proposed setup forms a low-overhead monitoring solution, with an average of 448 MB of memory and 6.8% CPU usage, while introducing 1.1s end-to- end latency for store operation and 0.6s for ledger querying respectively.

show abstract

“…Another crucial aspect focuses on the real-time utilization of node resources to schedule workloads [18]. This load-awareness is especially important in multi-tenancy cases [19], as interference effects such as cache misses and CPU context switches may lead to performance degradation. There are also papers which try to combine several of those aspects and propose a weighted multi-criteria decision strategy with the goal to optimize workload placement [20,21].…”

Section: Related Workmentioning

confidence: 99%

Enabling the Rescheduling of Containerized Workloads in an Ad Hoc Cross-Organizational Collaboration

et al. 2022

View full text Add to dashboard Cite

A group of organizations wishing to collaborate urgently, for example in case of a crisis, need to have a way to quickly deploy applications which enable them to speed up a potentially crisis-resolving decision-making process. A cross-organizational Kubernetes cluster, which is orchestrated by a central operator, allows to initiate these deployments in an ad hoc way. Performance issues may however arise at runtime, for example, a video pipeline belonging to a CCTV camera may produce a too low number of frames per second. The ad hoc cross-organizational collaboration case is especially prone to such issues as the set of candidate nodes and the environment in which they run may not be fully known to the operator. This article therefore motivates and describes the usage of a probe swarm architecture, which allows the operator to quickly generate an overview of the resource capabilities of a set of nodes, by executing code fragments locally. The obtained measurements can then enable the operator to decide on rescheduling operations. Evaluation of an illustrative probe swarm intervention shows that the performance of an example application could improve with factor five, ten or hundred when the pod would be rescheduled. This indicates that the proposed probe swarm architecture may complement other performance bottleneck detection techniques to improve performance of applications that need to be deployed urgently.

show abstract

Interference-Aware Orchestration in Kubernetes

Cited by 14 publications

References 19 publications

Object Detection in Driving Datasets Using a High-Performance Computing Platform: A Benchmark Study

Object Detection in Driving Datasets Using a High-Performance Computing Platform: A Benchmark Study

HLF-Kubed: Blockchain-Based Resource Monitoring for Edge Clusters

Enabling the Rescheduling of Containerized Workloads in an Ad Hoc Cross-Organizational Collaboration

Contact Info

Product

Resources

About