Improving Resource Utilization in Data Centers using an LSTM-based Prediction Model

Thonglek, Kundjanasith; Ichikawa, Kohei; Takahashi, Keichi; Iida, Hajimu; Nakasan, Chawanat

doi:10.1109/cluster.2019.8891022

Cited by 19 publications

(8 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Prasad and Madhuri [30] proposed a resource monitoring approach with reinforcement learning and machine learning concepts. Thonglek et al [9] used an LSTM model to accurately predict resource allocation for a given job. Two-layered LSTM discovers the trade-off between resource allocation and usage, and CPU and memory usage.…”

Section: Related Workmentioning

confidence: 99%

MVMS: RNN based Pro-Active Resource Scaling in Cloud Environment

Thakkar

Bhavsar

2023

SCPE

View full text Add to dashboard Cite

Cloud computing offers various services to its users, ranging from infrastructure, and system development environment, to software as a service over the internet. Having such promising services available over the internet consistently, it has become an ever-demanding facility. As a reliable services provider, a cloud service provider (CSP) needs to deliver its services seamlessly to users and is also required to optimally utilize the resources. Optimal resource utilization eliminates over and under-provisioning and improves the availability of cloud services. Therefore, it is a great need to have a model allowing CSP to systematize its resources to cater to customers' demands. Such a model should be computationally light and quick enough to produce effective results. In this work, a simple yet effective neural network-based resource prediction model named MVMS is proposed, which enables a CSP to predict the customer's resource demand in advance. The results show that compared to GRU, the proposed Multi-Variate Multi-Step (MVMS) model predicts the resources accurately. Thus, CSP can schedule the resources precisely and process real-time requests of users. Experiments on the bitbrains dataset indicate that the proposed MVMS resource prediction model is quick and accurate, with lower RMSE and MAE values.

show abstract

Section: Related Workmentioning

confidence: 99%

MVMS: RNN based Pro-Active Resource Scaling in Cloud Environment

Thakkar

Bhavsar

2023

SCPE

View full text Add to dashboard Cite

show abstract

“…Iqbal et al [18] proposed an adaptive observation window resizing method based on a 4-hidden-layer DNN for resource utilization estimation. The work of Thonglek et al [204] predicted the required resources for jobs by a two-layer LSTM network, which outperformed the traditional RNN model, with improvements of 10.71% and 47.36% in CPU and memory utilization, respectively.…”

Section: Resource Managementmentioning

confidence: 99%

Machine Learning Empowered Intelligent Data Center Networking: A Survey

Li¹,

Wang²,

Yang³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

“…In one approach, ML is used to solve scaling problems to predict threshold values dynamically and thereby provide application‐independent scaling 12,13 . In contrast, another method 14,15 uses a time‐series model to predict resource usage by reference to historical datasets. Supervised learning such as the random forest model has also been proposed for predictive autoscaling 16 .…”

Section: Related Workmentioning

confidence: 99%

RNN‐EdgeQL: An auto‐scaling and placement approach for SFC

et al. 2022

View full text Add to dashboard Cite

Summary This paper proposes a prediction‐based scaling and placement of service function chains (SFCs) to improve service level agreement (SLA) and reduce operation cost. We used a variant of recurrent neural network (RNN) called gated recurrent unit (GRU) for resource demand prediction. Then, considering these predictions, we built an intuitive scale in/out algorithm. We also developed an algorithm that applies Q‐Learning on Edge computing environment (EdgeQL) to place these scaled‐out VNFs in appropriate locations. The integrated algorithm that combines prediction, scaling, and placement are called RNN‐EdgeQL. RNN‐EdgeQL (v2) is further improved to achieve application agnostic group level elasticity in the chain, independent of applications installed on the VNFs. We tested our algorithm on two realistic temporal dynamic load models including Internet traffic (Abilene) and an application specific traffic (Wiki) on an OpenStack testbed. The contribution of this article is threefold. First, prediction model prepares the target SFC for the upcoming load. Second, an application agnostic characteristics of the algorithm achieves the group‐level elasticity in SFC. Finally, the EdgeQL placement model minimizes the end‐to‐end path of an SFC in multi‐access edge computing (MEC) environment. As a result, RNN‐EdgeQL (v2) gives the lowest overall latency, lowest SLA violations, and lowest VNFs requirement, compared to RNN‐EdgeQL (v1) and Threshold‐Openstack default placement.

show abstract

Improving Resource Utilization in Data Centers using an LSTM-based Prediction Model

Cited by 19 publications

References 10 publications

MVMS: RNN based Pro-Active Resource Scaling in Cloud Environment

MVMS: RNN based Pro-Active Resource Scaling in Cloud Environment

Machine Learning Empowered Intelligent Data Center Networking: A Survey

RNN‐EdgeQL: An auto‐scaling and placement approach for SFC

Contact Info

Product

Resources

About