Paper: The Next Generation of BGP Data Collection Platforms (Best Paper Award, SIGCOMM’24)
Authors: Thomas Alfroy, Thomas Holterbach, Thomas Krenc, KC Claffy, Cristel Pelsser
Presenter: Fabliha Labiba, Bangladesh University of Engineering and Technology, Bangladesh
Guest of Honor: Thomas Alfroy
Scribe: Aranya Saha
Time: 11am US ET / 5pm CSET / 11pm Beijing Time, Oct. 23, 2024
Zoom Link: Launch Meeting - Zoom
Are there some record videos uploaded？
Thank you to everyone who joined our first reading group session on The Next Generation of BGP Data Collection Platforms! Here is the link to the presentation slides for anyone interested in exploring the main insights and takeaways:
Presentation slides
QnA:
Q1(Jay Stewart): GILL seems very promising, but how can we encourage operators to deploy more vantage points (VPs)? Will operators be able to glean more insights from the data, or is it really only beneficial for network researchers in terms of data collection size reduction?
A1 (Thomas Alfroy): Regarding the second question – in fact, BGP data is very useful for network operators and not only for network researchers because whenever there is an issue with their connectivity for instance like - a failure or they are targeted by a hijack. They will have to process these BGP data to really understand what happened and troubleshoot the issue. The first part of the question on how to motivate network operators to pair with GILL: first of all, I think people are already motivated to pair with BGP data collection platforms. We discussed with route reuse operators, and they said they had to refuse some of the BGP vantage point pairing requests because of data management issues. We are also changing the way they set up the pairing session. For instance, they are currently just exchanging some emails with the BGP data collection platforms, so there is a lot of back and forth and it consumes a lot of human resources. This can demotivate some operators. So with GILL, we automated this process. If you go to our website, you can simply fill out a form where you input your ASN number, your IP address, and press enter; your BGP point is automatically deployed. By doing so, we think this can motivate them to set up a pairing session. Obviously, we thought about all the mechanisms to motivate them on our website, but obviously there is still some work to do on how to motivate them to set up a BGP session.
Q2 (Qiao Xiang): On one hand we want to encourage people to peer with skill or not even if not deal, but with ripe or route views all those BGP data collection platforms. On the other hand, people want to connect to this data platform is because we share the data we work together to identify issues or failures on the internet. But what about privacy? I think the January consensus people agree on beats like BGP is pretty good at hiding scenes. However, there are seminar papers on inferring business relationships between neighboring ASes, and we can already accurately infer these relationships. Suppose we had an infinite amount of BGP announcement data—if we had that, I think from a privacy perspective, we might even be able to reverse-engineer the configurations of BGP routers or autonomous systems. Is it possible? I am asking this because I once look into this and someone could infer accurate BGP configurations once these data platforms now. Would people be less open to share these things? What’s your opinion on this?
A2 (Thomas Alfroy): Obviously there might some network operators saying, okay so if I send you maybe you can reverse engineer traffic engineering setup or whatever other fancy things network operators can do. First of all, maybe, maybe we have lot of feeds you can reverse engineer. I am not an expert in internal BGP inference strategies but I think to do this, this will require a lot of feeds from the ASes to perform this accurately. If you only send one feed this is already super helpful for the platform but I don’t think this will be enough to reverse engineer what you are doing within your autonomous system. Second regarding BGP hijacks, because for instance someone can say that okay, if I am sharing my feed, I am maybe more exposed to BGP hijacks; this is something that I don’t think is true. Because BGP hijacks come from outside your network. So, if you are sharing you BGP routes, this will not help anyone perform a more efficient BGP hijack.
Q3 (Jay Stewart): When will your collector support IPv6?
A3 (Thomas Alfroy): this is just a prototype. It’s just running on a small VP that doesn’t have a IPv6 support. However you can send the IPv6 routes over IPv4 session. This capability is already enabled in the collector.
Q4 (Adrian Zapletal): 100% coverage would also not be needed, would it? Lots of tiny stub ASes don’t need a VP for us to get more information?
A4 (Thomas Alfroy): Every single AS in the internet has something unique, even if it’s 0.001%. Obviously there are some ASes that are more important. However, every AS in the internet is interesting, maybe useful at 10% or 0.001%. But every AS is useful.