Abstract
Recent research literature shows promising results by convolutional neural network- (CNN-) based approaches for estimation of traffic matrix of cloud networks using different architectures. Although conventionally, convolutional neural network-based approaches yield superior estimation; however, these rely on assumptions of availability of a large training dataset which is completely accurate and nonsparse. In real world, both these assumptions are problematic as training data size may be limited, and it is also prone to missing (or incomplete) measurements as well as may have measurement errors. Similarly, the 2-D training datasets derived from network topology based may be sparse. We investigate these challenges and develop a novel architecture which can cater for these challenges and deliver superior performance. Our approach shows promising results for traffic matrix estimation using convolutional neural network-based techniques in the presence of limited training data and outlier measurements.
Highlights
Internet traffic load is increasing multifold every few years
Two technologies are primary contributors for this increased traffic load: (1) cloud computing services which conveniently offer the Internet as a one-stop solution to all users in the form of infrastructure, platform, and software as service (AaaS, PaaS, and SaaS). (2) software-defined networks (SDNs) devise intelligent technologies to cater for ensuring Quality of Service (QoS) for Internet users even in presence of an ever-increasing Internet cross traffic
While the traditional flow level-based routing strategies to ensure QoS like MPLS, ATM, IntServ, and DiffServ failed in achieving scalability and QoS guarantees, SDNs employ programmable OpenFlow (OF) switches that communicate with SDN controllers via OF protocol to dynamically modify the traditional forwarding tables of routers based on a flowlevel control in scalable manner such as using hash-based approaches and employing more powerful hardware like multithreading controllers
Summary
Internet traffic load is increasing multifold every few years. In US alone, only the business internet traffic volume has seen a rapidly increasing trend growing from 45.1 billion GB in 2016 to 112.7 billion GB in 2020 and projected to increase up to 224.8 billion GB in 2023 [1]. (2) software-defined networks (SDNs) devise intelligent technologies to cater for ensuring Quality of Service (QoS) for Internet users even in presence of an ever-increasing Internet cross traffic. This places significant burden on the network as traditional traffic engineering (TE) techniques are effectively bypassed in software-defined networks. Recent research literature shows promising results by using neural network-based approaches for estimation of traffic matrix using different architectures, such as use of Artificial Neural Networks (ANNs) [11] and recurrent neural networks (RNNs) [12].
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have