The INITIATE project is a collaboration between leading universities in the UK, focusing on future Internet research. Lancaster University’s work includes large-scale experimentation with new technologies, interconnecting some of the leading 5G and next generation network testing facilities across the UK, allowing for dynamic interconnectivity amongst testbeds.
Through an extension of the MANY project, Lancaster University’s School of Communication and Computing has been able to utilise remote monitoring nodes in rural areas to test the behaviour of novel use cases in a real-world next generation wireless network environment. The demonstrator workshop allowed the MANY and INITIATE teams at Lancaster University to work in collaboration to deploy the monitoring and testing framework.
The MANY monitoring framework is divided into three layers:
- Application Layer monitoring focuses on Quality of Experience (QoE) metrics, measuring how the end user would holistically experience the network over time. For this the team streamed video media and determine a QoE score based on metrics such as video quality and buffering time.
- Network Layer monitoring utilises a variety of tools to actively determine the network throughput and latency over varying test conditions.
- Physical Layer monitoring examines the physical parameters recorded by the network infrastructure including, for wireless networks, aspects involving signal strength, background interference and channel noise.
Developed as part of MANY, the framework was used on the INITIATE testbed to determine end-to-end performance of the testbed under normal operating conditions.
There are some key differences from a typical MANY deployment, however, as all the deployed nodes are virtualised in remote locations, whereas a standard MANY installation will see Intel NUC devices deployed to the end node locations. The test deployment, also, uses a Virtual Machine located at both the University of Bristol and Lancaster sites. The Lancaster deployment acted as a server to where remote nodes test against, as well the destination for storing recoded metrics.
Two networks were present in the testbed, the first is the public internet – this was used for setup of the virtual machines, downloading the necessary packages from online repositories, as well as downloading the MANY monitoring framework from Docker Hub. The other was the physical private INITIATE fibre optic cable interconnecting Lancaster and Bristol via Slough. The performance testing was completed on this link, with metrics stored and plotted over time with the use of InfluxDB (a time series database) and Grafana to display these metrics.
Due to the transient nature of wireless networking, where network issues often appear for short periods of time (because of temporary interference or events such as very poor weather, which is typical in the location the project is working) the MANY networking stack is programmed to operate at very regular intervals. This tries to capture such events and allow researchers to examine the cause.
The results were in line with what is expected from the INITIATE network, which means future testing of multi-gigabit infrastructure within the MANY project will be as reliable as possible, with confidence that the performance recorded by the monitoring infrastructure is indeed what the end user is experiencing.