Building a framework to evaluate the impact of Inbound Traffic Engineering techniques
Put yourself in the shoes of an Internet Service Provider (ISP): your users are constantly fetching content from various sources, e.g. streaming services, social media and so on. Traffic from these sources enters your network over the inter-domain links connecting you to other Autonomous Systems (ASes). How much control do you have over the specific links that will be used and the amount of traffic going over them? Unlike cloud and content providers, who have a lot of control over their inbound traffic, ISPs have much fewer options. They can influence (e.g., using the MED or AS-path prepending) but they cannot really control or, at least, not without making compromises (e.g. such as reduced reliability and increased routing table sizes with selective prefix advertisements).
In our research, we have been developing a new technique that can aid ISPs achieve greater control over which links they receive traffic from. Our next step is to evaluate this technique with a realistic emulation. This is where your thesis comes in: its aim is to build a testbed that can mimic real world ISPs; emulate realistic traffic demands; prefix announcements and BGP policies; prefix assignments; and so on.
Milestones
- Using existing tools, set up a framework upon which you can test different traffic and routing scenarios from external ASes to your own.
- Add API hooks so that we may measure relevant metrics (e.g. traffic per user, per link…) as well as modify relevant parameters (e.g. DHCP config).
- Start measuring! Possibly refer to existing datasets, and analyse the control your ISP has on inbound traffic given various TE techniques.
Requirements
- Good programming skills in your language of choice.
- Knowledge of fundamental networking concepts such as BGP, IP prefixes, DHCP etc.
Supervisors
