Modelling the interactions between the Internet backbone and the BGP network Ivana Bachmann Felipe Espinoza NIC Labs, Universidad de Chile, Chile NIC Labs, Universidad de Chile, Chile ivana@niclabs.cl fdns@niclabs.cl 1 The Internet as an interdependent network system Abstract An interesting approach, is to pair together Given the importance of the Internet network the Internet Backbone and the BGP net- in our society, it is relevant to understand its work, in order to analyze a physical-logical behaviour under adverse scenarios. The In- network pair. However, to the best of our ternet can be studied through different an- knowledge, the articles applying interdepen- gles: by studying the Border Gateway Pro- dent networks models to study the Internet tocol (BGP) network, the Internet Backbone, robustness, have not paired these two net- the complete physical network, etc. However, works together [ZPC11, ATG16, ZZWY16, these networks do not exist in isolation, but WKVM16]. Thus, the purpose of this ongoing rather interact with one another (see figure study is to model the Internet as an interde- 1). Furthermore, the robustness behaviour of pendent network system composed by the In- interacting networks is different compared to ternet Backbone and the BGP network cou- their single network counterpart. In particu- pled together, and measure the Internet be- lar, it has been shown that networks can be haviour and robustness under adverse scenar- more fragile when coupled [BPP+ 10]. Indeed, ios, such as failures, or attacks. the single network approach to study the In- ternet’s behaviour has been criticized in the past by Willinger et al. [WR13] as it does not 2 Initial approach capture the whole behaviour of it. Thus, to We have previously presented an initial properly study the Internet we should model model [BBJ17]. Here, we modelled the In- it as an interdependent network system. ternet Backbone using a modified version of the relative neighborhood model [WKVM16], and the BGP network was modelled using a Scale-Free network with an appropriate λ value [FFF99] as it has been widely used to model BGP networks. In our modified ver- sion of the relative neighborhood model, each node is allocated into a 2-dimensional space, and any two nodes, u and v, get to be con- nected if there is no other node in the inter- section area of the circles centered at u and v, each of radius d(u, v), where d(u, v) is the eu- Figure 1: Interdependent networks example. Dotted clidean distance between node u and v. This lines represent interactions nodes of both networks. can be interpreted as follows: two nodes will get to be connected if there is no other node In: Proceedings of the IV School of Systems and Networks closer to them in-between them. (SSN 2018), Valdivia, Chile, October 29-31, 2018. Published at http://ceur-ws.org However, in this model the interconnections between both networks are established at ran- networks might be appropriate for modelling dom, and thus do not represent the actual net- the Internet Backbone and the BGP network. work pairing nature of the Internet. In or- However, we must note that the data used der to further develop this model, the relation here to represent the Internet Backbone corre- between Internet Backbone nodes, and BGP sponds to an approximation using the number nodes must be studied. Here, the hypothe- of countries in which a BGP node has physi- sis is that the number of Internet Backbone cal counterparts, and therefore does not show nodes interacting with a BGP node is propor- the real amount of Backbone nodes coupled tional to the degree of such BGP node. To to each BGP node. Thus, to accurately de- test this hypothesis, data has been collected to termine the relation between the amount of determine whether high degree nodes on the Internet Backbone nodes interacting with a BGP network are coupled to a proportional BGP node and the degree of said BGP node, number of nodes on the Internet backbone or we must determine the amount of Internet not. As an initial approximation of the Inter- Backbone nodes associated to each BGP node net Backbone, the localization of BGP nodes within each country. per country has been established. Thus, for the present work the hypothesis is that the number of countries in which BGP nodes have counterparts, is proportional to the degree of such node. 3 Data extraction The information to determine the country lo- calization of BGP nodes was obtained from the Routing Information Service (RIS) project from RIPE NCC [rip], and GeoLite2 geolocal- ization database [geo]. Here, the fAS prefixes obtained from the BGP routing tables were Figure 2: Each point represents a BGP node. For each used to determine their geographical localiza- node, we can see its degree versus the amount of coun- tion using GeoLite2. The router geolocaliza- tries in which that node has a physical counterpart. tion obtained from this kind of database has been demonstrated to be precise enough to perform localization analysis [GSH+ 17]. 5 Future work To obtain the degrees for BGP nodes, we used BGP routes obtained from RIS project and As future work, we will study again the re- Traceroutes from RIPE Atlas. lation between the BGP node degree and the amount of Backbone nodes connected to said We used the data obtained in this stage was BGP node, this time looking for a non-linear to get a first approximation about the relation relation. Also, further studies about the In- of the BGP nodes degrees and the amount of ternet Backbone, and the BGP network will physical counterparts of these nodes. be performed to continue the model develop- ment. In particular, this work will continue to 4 Results research on data to determine a more precise approximation of Internet Backbone nodes. From the data obtained we can observe that Once this work is completed, the possibility of there is no correlation between the amount adding other network infrastructures to the in- of countries in which a BGP node has physi- terdependent networks system model, that al- cal counterparts, and the degree of said node. low a better understanding of the Internet will Indeed, the Pearson correlation coefficient of be evaluated in order to improve the model. this data was of 0.23, showing the lack of cor- relation between the parameters studied. This can be further appreciated in figure 2. Acknowledgement These initial results show that in our initial This work was partially funded by CONICYT Doctor- model [BBJ17], the random coupling of both ado Nacional 21170165. References [ZPC11] Xian Zhang, Chris Phillips, and Xiuzhong Chen. An overlay mapping model for [ATG16] Abdulaziz Alashaikh, David Tipper, and achieving enhanced qos and resilience Teresa Gomes. Supporting differentiated performance. In Ultra Modern Telecom- resilience classes in multilayer networks. munications and Control Systems and In 2016 12th International Conference on Workshops (ICUMT), 2011 3rd Interna- the Design of Reliable Communication tional Congress on, pages 1–7. IEEE, Networks (DRCN), pages 31–38. IEEE, 2011. 2016. [ZZWY16] Qian Zhu, Zhiliang Zhu, Yifan Wang, and [BBJ17] Ivana Bachmann and Javier Bustos- Hai Yu. Fuzzy-information-based robust- Jiménez. Improving the chilean internet ness of interconnected networks against robustness: Increase the interdependen- attacks and failures. Physica A: Sta- cies or change the shape of the country? tistical Mechanics and its Applications, In International Workshop on Complex 458:194–203, 2016. Networks and their Applications, pages 646–657. Springer, 2017. [BPP+ 10] Sergey V Buldyrev, Roni Parshani, Ger- ald Paul, H Eugene Stanley, and Shlomo Havlin. Catastrophic cascade of fail- ures in interdependent networks. Nature, 464(7291):1025–1028, 2010. [FFF99] Michalis Faloutsos, Petros Faloutsos, and Christos Faloutsos. On power-law rela- tionships of the internet topology. In ACM SIGCOMM computer communica- tion review, volume 29, pages 251–262. ACM, 1999. [geo] Geolite2 geolocalization database. https://dev.maxmind.com/geoip/ geoip2/geolite2/. Accessed: 09-05- 2018. [GSH+ 17] Manaf Gharaibeh, Anant Shah, Bradley Huffaker, Han Zhang, Roya Ensafi, and Christos Papadopoulos. A look at router geolocation in public and commercial databases. In Proceedings of the 2017 Internet Measurement Conference, pages 463–469. ACM, 2017. [rip] Ripe probes. https://atlas.ripe.net/ probes/. Accessed: 09-05-2018. [WKVM16] Xiangrong Wang, Robert E Kooij, and Piet Van Mieghem. Modeling region- based interconnection for interdepen- dent networks. Physical Review E, 94(4):042315, 2016. [WR13] Walter Willinger and Matthew Roughan. Internet topology research redux. ACM SIGCOMM eBook: Recent Advances in Networking, 2013.