|Deliverable||Report / ACM IMC 2017 Paper / Master's Thesis, presentation at RIPE/NANOG/TNC|
As part of the RoN 2016 project we implemented an anycast testbed (http://www.anycast-testbed.nl). This work was presented at the plenary session of RIPE 73. As of today, this testbed is operational and it consists of 10 nodes worldwide – with 2 other nodes in the process of being deployed. We are now in the position to use the implemented testbed to study topics related to anycast operations. One of the topics of interest is the predictability of anycast ”preference choice”, and how choices are influenced by (BGP) route manipulation.
For RoN 2017, we will study the predictability of traffic patterns and anycast catchment (the routing of user to anycast site) even before an anycast site is deployed. To perform this study we will use traditional measurement tools, such as traceroute and RIPE Atlas, potentially combined with other data sources, such as BGP RIBs from RouteViews and RIPE RIS. We will also study how anycast catchments can be influenced by BGP route manipulation, such as route withdrawals or AS path prepending. We plan to provide insights into how much influence the operator of a given anycast network can have on the catchment, without having to contact its upstream providers.
Concretely, we will try to answer the following questions:
• Given two network locations, what will the division of traffic be if we deploy anycast instances there • How much can the catchment of an anycast network be influenced by AS Path prepending
• How can we apply this knowledge to optimize the resilience of anycast networks for DDoS attacks
Data and Tools
To make it possible for external parties to repeat our experiments, to use our results or to use our tools, we strive to make as much of the data that we gather and the tools that we develop publicly available.