Improving Africa’s Route Analysis.
Are you interested in making Africa’s Internet interconnections more transparent and insightful? Are you interested in helping to inform the community and network operators about the evolution of Internet routing and peering in Africa?
We are offering a funded PhD and a funded Masters position on the African Route-collectors Data Analyzer (ARDA) project. You will explore ways for improving the efficiency and effectiveness of route/peering data analysis and visualization. This project is to be undertaken in the Computer Science department at the University of Cape Town, with support from the Internet Society and Afrinic. Interested applicants should email me at (jchavula AT cs.uct.ac.za), providing a brief academic background and experience. Further details about the project are here .
I am also working on research projects described below. Prospective postgraduate students interested in any aspect of these projects are welcome to discuss with me.
Internet QoS Measurements and Network Engineering
Internet measurement platforms have become popular in recent years, and many of the platforms make their datasets publicly available (such as the M-Lab repository). Broad research questions to be tackled in this project include, firstly, on how best to orchestrate measurements for localized studies in low-resource areas, were there is also limited number of measurement probes ( such that there is need to combine data from different types measurement probes). Secondly, we explore how network measurements data can be used to improved network typologies. In this regard, we explore mechanisms for real-time analysis of network metrics in concert with SDN controllers in order to orchestrate traffic engineering. A third aspect of the research looks at how to improve visualization of Internet Measurements data so as to more readily inform the discourse on digital inequalities and policy formulation.
Open Research Projects
1) SDN Data-driven Internet Traffic Engineering
Emerging transport protocols such as MultiPath TCP (MPTCP) enable applications to take advantage of multiple network paths, thereby being able to achieve higher bandwidth and/or faster fail-over. In this project, we explore the use of SDN-based traffic engineering to help improve application performance in low-resource networks, such as in community networks. Overall, SDN controllers and nodes would incorporate and make use of the network measurements data (both user-centric and network-centric) to dynamically create high-performance overlays. Specifically, experiments are designed broadly as: 1) active measurements to test path support; 2) protocol performance optimization; 3) traffic offloading between Mobile Broadband and Wifi; 4) multipath performance measurements; 5) passive multipath performance measurements.
2) Pan-African Research and Education Networks
National Research and Education Networks (NRENs) continue to develop in Africa, among others, to provide Internet access to research institutions, run software and systems to allow scientific collaboration, and provide researchers with global access to digital research resources. Although many NRENs in Africa are still at developmental stage, there is already considerable and rapid interconnectivity aimed at to enabling continental and global collaboration. As the inter-NREN topologies are still developing, it is important to have mechanisms that could provide accurate information about performance to enable accurate analysis of the impact of different inter-connection strategies , and to support optimal allocation of network resources. This project seeks to explore and develop mechanisms for measuring performance and resource usage across pan-African NRENs, at very granular levels. This will include development and deployment of efficient measurement algorithms and infrastructure, as well as measurement orchestration platforms for carrying out thorough and long term analysis of the inter-NREN topologies.
3) User-Centric Network Measurements
A big focus of our research is on user-centric network measurements, and we are particularly interested in understanding how users, particularly in low-resource settings (such as in community networks, townships, public wifi) experience and perceive individual services and applications (e.g., video/audio streaming, gaming, conferencing, etc.) over different terminals (e.g., mobile phones, tablets, and computers), and over different networks and locations. Since these services have different requirements on performance and application-level metrics, their perceived performance need to be assessed separately.
As a use case, our first measurements study will form part of the iNethi Community Network, a localized content sharing and services platform currently being developed in Ocean View, a township located in the Southern Peninsula of Cape Town, South Africa. As part of the project, we will develop and deploy measurement probes, based on mobile phones and low-cost network devices (eg Raspberry Pi and Arduino) to facilitate mapping and measurement of the quality of service for various modes and locations of Internet access, including mobile networks, wifi hotspots, and fixed broadband. We carry out a longitudinal measurement and analysis of such metrics as: 1) device information; 2) application-oriented performance metrics for each type of services; 3) content information (e.g., codec, resolution, sending bitrate); 4) contextual information (location, time, etc.). In addition to the measurement, the studies will explore/integrate synthetic tools and surveys that should allow estimation of users’ perception regarding quality of experience for different services and network conditions.
4) Contextual/Geographical Visualization of Internet Measurements
From a user perspective, it is usually useful to be able to compare Internet access options within specific/localized geographical regions. Given that most Internet measurements data contains geo-location information, it is possible to design localized visualizations that should afford various communities the ability to view different Internet metrics at various levels of geographic/contextual granularity. Geographic visualizations would include availability quality of service for different networks, as well as cases of Internet censorship and Internet traffic manipulation.
In this project, we explore and visualize various types of local Internet datasets to help Internet communities better understand the state of connectivity in specific local jurisdictions (e.g. province, municipality, ward). Among others, the project will curate, aggregate and visualize Internet access and performance data obtainable through the Google BigQuery, OONI API, and Ripe Atlas among others.