Traffic profile analysis and suggestion for QoE improvement of a mobile application

OData support
Supervisor:
Dr. Varga Pál
Department of Telecommunications and Media Informatics

Nowadays the mobile network and its usage has become part of our everyday life, even its partial lack can cause dissatisfaction on ourselves. The mobile applicaton's market is still growing year by year such as the number of smartphone users. My thesis is focusing on the past few years’ biggest outcome: the Instagram application, where the average number of daily photosharing exeeds 55 million.

The majority of the most popular apps, as the example of Instagram shows as well, for the complete operation they require network access. The pattern of the network traffic generated by mobile apps tells us a lot about various circumstances. Analyzing the traffic on a package basis we can draw our conclusions regarding the following fields: user behaviour, the equipments used, the application and the actual characteristics of the network.

By structuring the network traffic profile of some application, relevant information regarding the QoE (quality of experience) will become on our disposal, furthermore certain application's use cases can be identified as well by its network pattern.

In my thesis I am going to expound in details the method that I used to analyze the network patterns related to Instagram. For the first step I defined the typical use cases of the application, which are as follows: timeline browsing, video wathching, glancing at profile pictures, etc.

The laboratory circumstances for the mesurements were mandatory factors in following the consistency conditions raised by myself. After the validated network measurements at first I made a detection structure in an intuitive manner, then followed by the analysis and the comparison of certain group cases, I specified them as well. Even Instagram API played an important role in the process, since its script helped me download small- and large size pictures and videos as well with their metadata included.

A parameterized network analyzer software was structured then in order to classify automatically the user groups. The software categorizeses the behaviour of the users connected to the measured network considering the following factors: the list of domains used by Instagram, the characteristics of TLS handshake in case of encrypted traffic and the momentary transmission speed. The study and the method give opportunity to the developers to progress and to parameterize easily and quickly the dynamic network management, through which the QoE can be increased in an application-specific way.

The user satisfaction is greatly influenced by the download time of certain contents and by the limited bandwidth. Considering the user demand as the last subject of my thesis I define the Instagram-specific performance indicators, which provide further support for the development of the SDN's dynamic network management.

Downloads

Please sign in to download the files of this thesis.