Hybrid deep learning models and link probability based routing in software defined-DCN

In an SDN-based Data Center Network (SD-DCN) environment, network traffic is highly vulnerable to handle the dynamic requirements of the users. It creates difficulty for network management tasks and future network routing path prediction. Deep learning is an emerging technology available to analyze...

Ausführliche Beschreibung

Gespeichert in:
Bibliographische Detailangaben
Veröffentlicht in:The Journal of supercomputing Jg. 79; H. 9; S. 9771 - 9794
Hauptverfasser: Modi, Tejas M., Swain, Pravati
Format: Journal Article
Sprache:Englisch
Veröffentlicht: New York Springer US 01.06.2023
Springer Nature B.V
Schlagworte:
Algorithms
Artificial neural networks
Communication
Communications traffic
Compilers
Computer Science
Datasets
Deep learning
Delay
Error analysis
Interpreters
Machine learning
Mathematical models
Network topologies
Neural networks
Path predictors
Processor Architectures
Programming Languages
Quality of service architectures
Recurrent neural networks
Simulation
Software
Software defined-data center network (SD-DCN)
Long short term memory (LSTM)
Bidirectional long short-term memory (BiLSTM)
Hybrid model
Convolution neural network (CNN)
ISSN:0920-8542, 1573-0484
Online-Zugang:Volltext
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
Beschreibung
Zusammenfassung:In an SDN-based Data Center Network (SD-DCN) environment, network traffic is highly vulnerable to handle the dynamic requirements of the users. It creates difficulty for network management tasks and future network routing path prediction. Deep learning is an emerging technology available to analyze the network environment to enhance network performance. The advanced deep learning models can easily handle dynamic changes as compared to conventional mechanisms in the network environment. The Convolution Neural Network (CNN) deep learning model provides limited congestion features extraction. The Recurrent Neural Network (RNN) deep learning model provides long feature dependency information from the given traffic dataset. To overcome the limitations of CNN and RNN models, an efficient deep learning technique is required to find both congestion features and feature dependency from previous input data. Moreover, in this paper, the mathematical model is proposed which improves network routing and flow management mechanisms by using available link probability and computing average network delay in the network. The proposed mathematical model provides the average network delay and traffic dataset to CNN-RNN hybrid deep learning models to analyze the network traffic. The improvement of the Quality of Service (QoS) in the network is the major aim to incorporate a mathematical model in the proposed system. In the proposed work, the hybrid deep learning models are a combination of two deep learning models, i.e., CNN with Long Short Term Memory (LSTM) network as CNN-LSTM and Bidirectional Long Short Term Memory (BiLSTM) network as CNN-BiLSTM. The experimental results show that the proposed approach outperforms the existing work in terms of several metrics such as average network throughput, average network delay, packet loss rate, and routing overhead. Furthermore, the result comparison is analyzed for the training and testing error analysis of different deep learning models.
Bibliographie:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0920-8542
1573-0484
DOI:10.1007/s11227-022-04995-2