Abstract
Network utilisation efficiency can, at least in principle, often be improved by dynamically re-configuring routing policies to better distribute ongoing large data transfers. Unfortunately, the information necessary to decide on an appropriate reconfiguration—details of on-going and upcoming data transfers such as their source and destination and, most importantly, their volume and duration—is usually lacking. Fortunately, the increased use of scheduled transfer services, such as FTS, makes it possible to collect the necessary information. However, the mere detection and characterisation of larger transfers is not sufficient to predict with confidence the likelihood a network link will become overloaded. In this paper we present the use of LSTM-based models (CNN-LSTM and Conv-LSTM) to effiectively estimate future network traffic and so provide a solid basis for formulating a sensible network configuration plan.
Highlights
The evolution of data engineering has increased the frequency of use of applications and data transmission services
The CNN, long short-term memory (LSTM) and CNN-LSTM models are trained using the following set of parameters: the throughput (TH), the number of active files for a given time step (AF), i.e. the number of files being transported by FTS; the queue length (QL), or number of submitted files (SF) and, the average active files size (AFS), that were processed since the last FTS report
The first task describes the impact of transfers on network traffic at a given moment in time for Γ ≥ 0, and it is evaluated by measuring the MSE0 in data set representing transfers from TRIUMF
Summary
The evolution of data engineering has increased the frequency of use of applications and data transmission services. It is not uncommon to observe link saturation during data transfer, when network traffic occupies almost all the available bandwidth. The possibility to dynamically configure the network to add bandwidth and load-balance traffic could reduce occurrences of link saturation and so improve transfer performance and prevent recoverable (server or networking) errors [2]. In this context, it becomes essential to design a Deep Learning model for traffic estimation [4], ensuring a flexible and adaptive learning process. An exception is the occurence of sudden throughput drops caused by unexpected errors Most of these are quickly rectified by FTS so that when observing the network traffic we do not see a long-term problem. The paper is organised as follows: the section presents the data sets, a description of the data pre-processing step and the DL models follows, together with an analysis of the results leading to the choice of the best model
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
