| Summary: | Network Orchestrators are the brains of 5G networks. The orchestrator is responsible for the orchestration and management of Network Function Virtualisation Infrastructure (NFVI), understanding network services on NFVI and software resources. The International Telecommunication Union (ITU) have categorized three main 5G network services for the orchestration. So called, Enhanced Mobile Broadband (eMBB), Ultra-reliable and Low-latency Communications (uRLLC) and Massive Machine Type Communications (mMTC). Categorizing the network is achieved in 5G by a method called network slicing. In the future, a device connecting to a 5G network will be in one of three slices (eMBB, uRLLC and mMTC) based on network characteristics. The focus of this dissertation goes to the eMBB slice. Normally day-today internet users will use the eMBB slices. Thus, all the daily internet access such as watching YouTube videos, making Skype video calls, calling via WhatsApp, downloading files, listening to online radio and whatnot will happen via eMBB slice. However, this approach neglects the importance of the web application a user is using in the eMBB slice. For example, a family doctor may give first aid assistance via a Skype video call in an emergency situation. Thus the call of the doctor, in this case, should be prioritized over other normal daily web tasks. Thus, there is a requirement of prioritizing usual web-tasks in certain scenarios which eMBB slice neglects. It is possible to detect websites or web plications with modern-day technologies. Hence, these type of website detection algorithms can be improved to detect web-tasks (Skype voice calling, Skype video calling, etc...) to provide a separate slice within eMBB slices upon doctor's request. The goal of this study is to identifying web-tasks by capturing the network data packets flowing in and out of the system and perform an application-based classification by using machine learning techniques. After the classification, data was fed to the 5G Orchestrator or to the 5G Core. The Orchestrator will allocate a number of Network Function Virtual Machines to provide best quality of service (QoS) based on generated slice information. iv In this research, a Website Task Finger Printing (WTFP) algorithm is introduced to identify web traffic (such as identifying if a user is watching a video on Facebook, rather than just detecting the website that they are viewing). Possible applications of the developed algorithm vary from 5G ultra slicing to network security. This study delves deeper into Website Finger Printing (WFP). Traditional papers only describe how to identify websites by using statistical analysis, whereas this study shows how we can identify what task a user is performing rather than just which website they are currently visiting. The identifier captures the inbound and outbound data and then uses the packet length histogram as the main feature. After that, application-based features were extracted by using heuristic logical filters to prepare a feature vector for the Machine Learning (ML) algorithm. A trained Multi-layer Perceptron (MLP) based Artificial Neural Network (ANN) was selected as the classifier after comparing results with Support Vector Machine (SVM), Recurrent Neural Network (RNN) and Convolutional Neural Network (CNN). The MLP algorithm was able to classify website tasks with 95.50% accuracy. After classification, the classified class was sent to the 5G Orchestrator, then it refers to programed Network Service Descriptor and based on our specifications generates a new slice by using Network Slice Engine (NSE). After that, it monitorsthe present bitrate of the slice by using Zabbix. Next, the Orchestrator either increase or decrease the bitrate to give the optimum Quality of Service (QoS) by using Auto Scaling Engine (ASE). The algorithm also used to generate specific QoS by using Open5G Core. Therefore, this study shows that it is possible to allocate slices based on webtasks in 5G Mobile network thus proposing to investigate further; to enable web-task based slicing for the future mobile networks.
|
|---|
