Comparison of Machine Learning Techniques for VNF Resource Requirements Prediction in NFV

Moradi, Mahsa; Ahmadi, Mahmood; Nikbazm, Rojia

doi:10.1007/s10922-021-09629-1

Cited by 13 publications

(9 citation statements)

References 19 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This involves the development and evaluation of algorithms that intend to predict and address the network's resources needs dynamically by deploying VNF instances over the virtualized infrastructure. The works given in [23], [24], [25] and [26] provide a model of these solutions.…”

Section: State Of the Artmentioning

confidence: 99%

Noisy Neighbour Impact Assessment and Prevention in Virtualized Mobile Networks

Muro¹,

Baena²,

Fortes³

et al. 2023

IEEE Trans. Netw. Serv. Manage.

View full text Add to dashboard Cite

The generalization in the use of virtualization in the upcoming generation of cellular networks involves new paradigms and approaches for their management. The correct sharing of the underlying resources between multiple virtualized network functions as well as any other processes sharing the same computational platform implies a complex architecture that makes virtualization challenging. As a result, numerous variables (e.g., computational capacity) were mostly ignored by previous management systems, but that can now lead to relevant impacts on the service performance. In this virtualized scenario with multiple coexistent processes over the same hardware, a "Noisy Neighbour" (NN) is identified as an entity that uses most of the underlying resources while other virtual units suffer a lack of them. While difficult to identify, such situations can affect the network service. In this context, the present work analyzes and assesses the NN problem for 5G Core scenarios. A complete emulated 5G network and analysis framework is defined and developed to evaluate the impact of a noisy entity. In this way, the degradation that Key Performance Indicators suffer in the network and by the end-users when a NN appears is assessed. Thus, the present work proposes a baseline for handling NN through a novel lifecycle management flow. For this purpose, it is evaluated the effectiveness of multiple Machine Learning (ML) models for the identification of NN, based on the metrics gathered from the proposed framework, achieving 99% of accuracy. Moreover, ML is applied to develop a method for network performance inference, along with a prediction model to forecast the number of CPU resources the network may demand at any given time supporting the proposed management flow.

show abstract

Section: State Of the Artmentioning

confidence: 99%

Noisy Neighbour Impact Assessment and Prevention in Virtualized Mobile Networks

Muro¹,

Baena²,

Fortes³

et al. 2023

IEEE Trans. Netw. Serv. Manage.

View full text Add to dashboard Cite

show abstract

“…Proposal of numerous NFV and mobile networking management mechanisms leveraging state-of-the-art machine learning techniques has garnered huge interest in recent years throughout the research community. 2,3,[5][6][7][8][9][10][11][12][13][14][15] Despite this enthusiasm, deep knowledge of several aspects of NFV as a whole (MANO, services, overhead reduction, latency minimization, etc.) and of the potential, but also of the limitations of machine learning techniques are crucial to designing robust, simple and valuable solutions to the administrators of NFVIs.…”

Section: Ml-based Vnf Resource Usage Predictionmentioning

confidence: 99%

“…Therefore, the resulting resource usage forecasts of many time steps of different resource attributes from an SFC provide high-accuracy, high-fidelity predictions since they benefit from variations in resource loads of multiple resource attributes that are highly correlated to those targeted resource attributes that we aim to forecast. For instance, some traditional time series-based prediction approaches like auto-regressive integrated moving average (ARIMA) enhanced with trend and seasonality filtering, 5 decision tree (DT) and support vector regression, 13 Markov decision process, 15 and softmax regression 12 are still popular because of their effectiveness, swiftness and overall simplicity at predicting future resource usage of non-stationary, univariate resource attributes like CPU, memory, disk I/O, or network traffic. However, since these approaches leverage univariate (i.e., take historical data from a single resource attribute to predict a resource usage horizon of that same attribute) resource usage history to forecast a resource usage horizon, none of them can reinforce their predictions from resource usage inter-dependencies from multiple correlated sources.…”

Section: Ml-based Vnf Resource Usage Predictionmentioning

confidence: 99%

“…Some useful techniques to select input data features prior to ML model training to remove uncorrelated data while avoiding reducing the models' prediction accuracy can be found in the literature. One of those is applying a GNN to a directed graph 2,13,19,20 and the other is feature pruning using correlation coefficients. [21][22][23][24] GNN theory 20 is a popular choice among teams designing ML-based solutions in which directed graphs are involved.…”

Section: Automated Input Feature Selection Techniquesmentioning

confidence: 99%

“…Proposal of numerous NFV and mobile networking management mechanisms leveraging state‐of‐the‐art machine learning techniques has garnered huge interest in recent years throughout the research community 2,3,5‐15 . Despite this enthusiasm, deep knowledge of several aspects of NFV as a whole (MANO, services, overhead reduction, latency minimization, etc.)…”

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

NFVLearn: A multi‐resource, long short‐term memory‐based virtual network function resource usage prediction architecture

2022

View full text Add to dashboard Cite

Virtual resource load prediction in network function virtualization (NFV) is the subject of intense research due to its crucial role in enabling proactive resource adaptation in dynamic NFV environments whose resource demand constantly changes. Several long short‐term memory (LSTM)‐based approaches have been proposed to forecast the resource load of multiple resource attributes of a virtual network function (VNF) in a service function chain (SFC). In this article, we present NFVLearn, a flexible multivariate, many‐to‐many LSTM‐based model which uses different types of resource load history (CPU, memory, I/O bandwidth) from various VNFs of an SFC to predict future loads of multiple resources of a VNF. We then compare four novel automated input selection frameworks for NFVLearn. Simulations on those frameworks based on graph neural networks, Pearson correlation coefficient, Spearman rank correlation coefficient, and Kendall rank correlation coefficient demonstrate that models using lesser, highly correlated input features retain high prediction root mean squared error accuracy and coefficients of determination scores by leveraging resource attribute inter‐dependencies from the SFC. Those results show that resource attribute interdependency‐based input feature selection frameworks can reduce overhead in the control plane while keeping high accuracy and high fidelity resource load prediction of multiple resource attributes.

show abstract

Value is King: The MECForge Deep Reinforcement Learning Solution for Resource Management in 5G and Beyond

et al. 2022

View full text Add to dashboard Cite

Multi-access edge computing (MEC) is a key enabler to fulfill the promises of a new generation of immersive and low-latency services in 5G and Beyond networks. MEC represents a defining function of 5G, offering significant computational power at a reduced latency, allowing to augment the capabilities of user equipments while preserving their battery life. However, the demands generated by a plethora of innovative and concurrent IT services requiring high quality of service and quality of experience levels will likely overwhelm the—albeit considerable—resources available in 5G and Beyond scenarios. To take full advantage of its potential, MEC needs to be paired with innovative resource management solutions capable of effectively addressing the highly dynamic aspects of the scenario and of properly considering the heterogeneous and ever-changing nature of next generation IT services, prioritizing the assignment of resources in a highly dynamic and contextual fashion. This calls for the adoption of Artificial Intelligence based tools, implementing self-* approaches capable of learning the best resource management strategy to adapt to the ever changing conditions. In this paper, we present MECForge, a novel solution based on deep reinforcement learning that considers the maximization of total value-of-information delivered to end-user as a coherent and comprehensive resource management criterion. The experimental evaluation we conducted in a simulated but realistic environment shows how the Deep Q-Network based algorithm implemented by MECForge is capable of learning effective autonomous resource management policies that allocate service components to maximize the overall value delivered to the end-users.

show abstract

Comparison of Machine Learning Techniques for VNF Resource Requirements Prediction in NFV

Cited by 13 publications

References 19 publications

Noisy Neighbour Impact Assessment and Prevention in Virtualized Mobile Networks

Noisy Neighbour Impact Assessment and Prevention in Virtualized Mobile Networks

NFVLearn: A multi‐resource, long short‐term memory‐based virtual network function resource usage prediction architecture

Value is King: The MECForge Deep Reinforcement Learning Solution for Resource Management in 5G and Beyond

Contact Info

Product

Resources

About