Lehrstuhl für Informatik III


Cloud Applications and Networks (CAN)

A fundamental part of any application or service today is the ability to communicate. Each service, each application uses some form of communication to, for example, operate resource-efficient in the cloud or simply to connect users, or load and display data from remote servers from the Internet. The applications make use of a huge pool of possibilities how they communicate. This includes various access technologies such as 5G mobile communications or wireless lan, or different protocols for downloading data such as streaming. Furthermore, economic interests and cost pressure, but also the technical possibilities offered by cloud architectures, enable the implementation of a service or application in scalable form. Nowadays, applications are not simply created, but use possibilities to serve any number of users. So a service can be rolled out in the cloud if necessary, so that it is available for 10, 1000 or one million users with consistent quality.

The objective of the Cloud Applications and Networks working group is the evaluation, modeling and performance analysis of Internet applications in relation to their communication technologies in the different communication networks. The working group always carries out the steps monitoring, modeling and analysis from a scientific point of view and with scientific methods. The monitoring of complex cloud infrastructures with dynamic Internet applications on mobiles, in the Internet or in a data center is an integral part of this process, so that targeted information can be gained, the complex interrelations can be understood and conclusions can be drawn. Based on the information, analyzes can then be compiled using abstract models with simulations if necessary, which has been the core competency of the entire "Chair of Communications Networks" for years.

Modeling of Internet applications leads to an abstract understanding of the operation of an Internet application or service or a form of communication. Abstract models allow easy use in network and application studies for insights, troubleshooting, optimizations, enhancements, and scaling and overload control methods, which are some of the main activities of the group for Cloud Applications and Networks.

Quality of Experience allows us to view and evaluate an application in terms of user experience. Technical data fades into the background, as long as it does not directly affect the user and her/his application experience. The Quality of Experience allows to base on the essential factors and perform targeted analyzes on the user of the application.

Monitoring is essential for all analyzes of applications and networks. It serves to collect and gather the relevant information for conclusions. In the course of increasingly complex and demanding applications on a variety of communication channels, an intelligent monitoring is needed, which manifests a scientific research focus of the group. It often has to be measured in various system layers, such as at the user, in the network or on the Internet because of technical limitations or accuracy of the data. The research question here is how to implement efficient monitoring for specific applications and how it can be carried out under certain constraints and conditions.

Research Questions

Methodology: What are the techniques and methods for modeling today's Internet services?
(especially with regard to today's major services on the Internet such as YouTube or emerging use cases like Internet of Things)

Structure and composition of applications: How does a large-scale Internet service look like today?
What are the common key elements, common paradigms and methods that can be identified as performance critical?

Communication protocols and access: What are and how do communications paradigms and protocols meet user expectations or contribute to application performance?

How to satisfy users? What interests a user in an Internet or mobile application? What is crucial for the user?

How should my application be structured to meet user expectations?

What are the fundamental relationships between network performance and user satisfaction?
What are tolerated network properties and when do users feel disturbed?

Which properties does a monitoring have to fulfill?
How to build a targeted monitoring for a variety of purposes?
Where is the degree between useful information for scalability and modeling, and useless, unanalyzable data overhead through monitoring?
How do I create efficient monitoring for purposeful data analysis that includes both application and network information?

Large-scale and comprehensive monitoring:
How to create a large-scale spatially inclusive and comprehensive performance monitoring system for mobile networks?
How can I establish a comprehensive, all-encompassing monitoring that includes the user and her/his experience on all devices?

Crowdsourced monitoring:
How to implement a crowdsourced-based rating/crowdsourced monitoring?
How do I interpret user ratings? What challenges arise?

Location and type of monitoring:
How to monitor a complex, large ecosystem of today's Internet applications with user and (mobile) network to allow meaningful conclusions from data?

Continuous evaluation:
How to create a continuous evaluation of systems on the Internet? How to create a large-scale evaluation system? What challenges arise for today's systems?


Dr. Stefan Geißler

Head of Research Group

Anika Seufert M. Sc.
(née Schwind)

Doctoral Researcher

Viktoria Vomhoff M. Sc.

Doctoral Researcher

Simon Raffeck M. Sc.

Doctoral Researcher

Marvin Ewald B.Sc.

Student Researcher

Julian Kargl B.Sc.

Student Teaching Assistent

Thomas Meier

Student Teaching Assistent

Yannick Pfeiffer B.Sc.

Student Researcher

Marleen Sichermann B.Sc.

Student Researcher

Filip Simonovski B.Sc.

Student Teaching Assistent

Minh Duc Tran B.Sc.

Student Researcher

Maximilian Ziegler

Student Researcher

Current Projects

5CALE - Massive scaling of fully virtualized 5G mobile core networks in the context of IoT
(Februray 2019 - January 2022)

Within the project, new scaling methods and resource management approaches are being developed for the next generation 5G mobile communication network with IoT traffic. It is funded within the 5G call "Digitale Offensive" of the Bavarian Ministry of Economic Affairs with a total budget of one million euros and a term of 3 years.

5GQMON - Crowdsourcing-based measurement methodology of user-oriented quality of service and network quality in 5G mobile networks
(May 2021 - April 2024)

The aim of the project is to investigate and define reliable crowdsourcing and agent-based methods to measure the performance of the new 5G mobile networks and their new applications, such as industrial applications, smart city, and the Internet of Things (IoT), according to end user's quality of experience.

Tools and Datasets

WhatsAnalyzer is a web-based tool to collect and analyze chat histories of the mobile messaging application WhatsApp. With the help of this tool, we analyze the group communication behavior in WhatsApp and investigate possible implications of this emerging communication paradigm on networking technology. 
[ More | Contact ]

YoMo-Docker is a Docker container to actively measure QoE related factors of YouTube video streaming. The measurement concept is based on emulating a virtual end-user device requesting video streams, which are then monitored at the network and application layers, on the basis of QoE-relevant features.
[ GitHub | Contact ]

YomoApp (YouTube Monitoring App) is an Android app. With it you can watch YouTube videos and rate the quality of the streaming. On a map, you can see how well your own network and other networks are working in different locations. So you can easily compare different providers and can see directly where the mobile network works well and where it has to be improved. Developers can download data about the streaming performance.
[ More | Contact ]

Wrapper App is a measurement application to measure the native YouTube app for smartphones from the Google Play Store. Among other things, the current quality and playing time is read from the app and written in a log file.
[ Github | Contact ]

KOMon is a tool developed to characterize and monitor the packet processing times of softwareized network functions by performing in-stack monitoring. In-stack monitoring leverages the network stack of the platform it is deployed on to monitor packets with high accuracy and low overhead.
[ Github | Contact ]

JTableVisor is a transparent proxy-layer for the OpenFlow control channel. It enables a flexible and scalable abstraction of multiple data plane devices into one emulated data plane switch, meeting the requirements of the control plane applications. Therefor, TableVisor registers with the SDN controller as a single switch with use-case specific capabilities. It translates instructions and rules from control applications towards the appropriate data plane device where they are executed.
[ Github | Contact ]

QoE3 (qoecube) is a dataset for YouTube's mobile streaming client, which follows the popular Dynamic Adaptive Streaming over HTTP (DASH) standard. The data was measured over 4 months, at 2 separate locations in Europe, at the network, transport and application layer for DASH.
[ More | Contact ]

VideoMon-IJNM2018 Leveraging the Measuring Mobile Broadband Networks in Europe (MONROE) test bed that enables experimentation with 13 different network configurations in four countries, we collect more than 2114 measurement samples in operational mobile networks. Per sample we collected, among other things, the available bandwidth, signal strength as well as application layer video QoS data. 
[ Download | Contact ]

Initial Delay Dataset YouTube is an extensive dataset to study the response time of YouTube's mobile video streaming service on Android. The dataset consists of time-synchronized log files from the network, transport, and application layer. The measurements were performed from November 23, 2017 to February 05, 2019 for 75 network scenarios and include 4498 runs with 783.88 hours of video playback time.
[ Download | Contact ]