Boundaries of Flow Table Usage Reduction Algorithms Based on Elephant Flow Detection

Jurkiewicz, Piotr

doi:10.23919/ifipnetworking52078.2021.9472832

Cited by 7 publications

(4 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The benchmark for assessment is the theoretical flow table occupancy reduction rate curves of a model that can perfectly predict the size of all flows from their first packet. This approach is detailed in [30] as the first method. The process involves selecting the smallest number of largest flows, ordered by size in descending order, which cumulatively accounts for a predetermined percentage of total network traffic.…”

Section: Resultsmentioning

confidence: 99%

Elephant Flow Classification on the First Packet With Neural Networks

Kądziołka,

Jurkiewicz,

Wójcik

et al. 2024

IEEE Access

Self Cite

View full text Add to dashboard Cite

Quick and accurate identification of the largest flows in the network would allow for the management of most traffic using dedicated, flow-specific routes and policies, thereby significantly reducing the overall number of entries in switch flow tables. Our analysis focuses on utilizing neural networks to classify elephant flows based on the first packet using 5-tuple packet header fields. The findings indicate that with simple neural networks comprising solely linear layers, it is possible to accurately detect elephant flows at their inception, thereby reducing the number of flow table entries -by up to a factor of 15 -while still effectively covering 80% of the network traffic with individual flow entries.

show abstract

Section: Resultsmentioning

confidence: 99%

Elephant Flow Classification on the First Packet With Neural Networks

Kądziołka,

Jurkiewicz,

Wójcik

et al. 2024

IEEE Access

Self Cite

View full text Add to dashboard Cite

show abstract

“…This projection is predicated on the assumption of perfect prediction of each flow's size on its initial packet. This methodology, described in [35] as the first method, involves selecting the smallest subset of the largest flows, arranged by size in descending order, which collectively represent a predetermined percentage of the total network traffic.…”

Section: Flow Table Reduction Vs Traffic Coveragementioning

confidence: 99%

Utilizing TabNet Deep Learning for Elephant Flow Detection by Analyzing Information in First Packet Headers

Kądziołka,

Jurkiewicz,

Wójcik

et al. 2024

Entropy

View full text Add to dashboard Cite

Rapid and precise detection of significant data streams within a network is crucial for efficient traffic management. This study leverages the TabNet deep learning architecture to identify large-scale flows, known as elephant flows, by analyzing the information in the 5-tuple fields of the initial packet header. The results demonstrate that employing a TabNet model can accurately identify elephant flows right at the start of the flow and makes it possible to reduce the number of flow table entries by up to 20 times while still effectively managing 80% of the network traffic through individual flow entries. The model was trained and tested on a comprehensive dataset from a campus network, demonstrating its robustness and potential applicability to varied network environments.

show abstract

“…The phenomenon of elephant and mice flows can lead to poor network performance [2]. Limitations in storage and processing resources make it impossible to collect and monitor all network packets.…”

Section: A Backgroundmentioning

confidence: 99%

“…These flows are called elephant flows. The remaining flows, which are large in number but carry very little traffic, are called mice flows [2]. The latter carry most traffic in bytes, but most flows are mice.…”

Section: Introductionmentioning

confidence: 99%

Traffic Classification in IP Networks Through Machine Learning Techniques in Final Systems

Gómez¹,

Riaño²,

Ramírez-González

2023

IEEE Access

View full text Add to dashboard Cite

Data centers in higher education institutions, as well as those of large corporations, face challenges in terms of traffic flow management. In some cases, due to the limited hardware resources used for this purpose, and in others, despite having enough high-performance equipment, the centers lag behind when the traffic flow grows exponentially due to the memory limitations of the devices, which slows down the network performance. The contribution of this investigation work is the implementation of a classifying elephant and mice system using machine learning techniques for the early detection with the first flow based on the dynamic calculation of the threshold, according to the input parameters of the final system. In the first instance, training algorithms are used to determine the best performance, then the proposed algorithm determines the model with the best prediction, obtained from the supervised learning algorithm trained in off line mode. Finally in the phase of online prediction, the algorithm is capable of predicting with high precision the type of traffic in terms of the input flow, and updates in a dynamic way the threshold to determine whether the traffic is elephant or mice. With this information the network hardware can decide then to route the flows according to their characterization. According to the results, the model that best generates predictions is the decision tree with a 100% confidence level.

show abstract

Boundaries of Flow Table Usage Reduction Algorithms Based on Elephant Flow Detection

Cited by 7 publications

References 34 publications

Elephant Flow Classification on the First Packet With Neural Networks

Elephant Flow Classification on the First Packet With Neural Networks

Utilizing TabNet Deep Learning for Elephant Flow Detection by Analyzing Information in First Packet Headers

Traffic Classification in IP Networks Through Machine Learning Techniques in Final Systems

Contact Info

Product

Resources

About