Internet Routing COS 598 A Today Interdomain Topology

Internet Routing (COS 598 A) Today: Interdomain Topology Jennifer Rexford http: //www. cs. princeton. edu/~jrex/teaching/spring 2005 Tuesdays/Thursdays 11: 00 am-12: 20 pm

Outline • Interdomain topology – AS graph – Inferring the topology from routing data – Structure of the AS graph • AS relationships – Common pair-wise relationships – Inferring the relationships from routing data – Characteristics of the relationship graph • Peering policies – Example of AOL’s peering agreement

What is the AS Graph? • Node: Autonomous System • Edge: Two ASes that speak BGP to each other 4 3 5 2 1 7 6

What is an Edge, Really? • Edge in the AS graph – At least one BGP session between two ASes – Some destinations reached from one AS via the other d d AS 1 Exchange Point AS 2 AS 3

How Do You Know a Node or Edge Exists? • Consult the Whois database? – Tells which ASes have been allocated – But, might be out-of-date on who owns it – … and often doesn’t say who the neighbors are • See a path that uses the node/edge – Collect measurements of AS paths – Extract all of the nodes and edges – E. g. , AS path “ 7018 1 88” implies • Nodes: 7018, 1, and 88 • Edges: (7018, 1) and (1, 88)

Interdomain Routing: Border Gateway Protocol • ASes exchange info about who they can reach – IP prefix: block of destination IP addresses – AS path: sequence of ASes along the path • Example: a BGP route in AS 7018 shows – Prefix 12. 34. 158. 0/24 has path “ 7018, 1, 88” “ 12. 34. 158. 0/24: path (88)” “ 12. 34. 158. 0/24: path (7018, 1, 88)” 7018 88 1 data traffic 12. 34. 158. 5

Where to Get BGP Routes: Public Servers 4 7018 1221 701 7 80 3786 9. 184. 112. 0/20 3. 0. 0. 0/8 BGP sessions

Example: BGP Table Network * 3. 0. 0. 0/8 * * *> * * 9. 184. 112. 0/20 * *> * * * (“show ip bgp” at Route. Views) Next Hop Metric Loc. Prf Weight Path 205. 215. 45. 50 0 4006 701 80 i 167. 142. 3. 6 0 5056 701 80 i 157. 22. 9. 7 0 715 1 701 80 i 195. 219. 96. 239 0 8297 6453 701 80 i 195. 211. 29. 254 0 5409 6667 6427 3356 701 80 i 12. 127. 0. 249 0 7018 701 80 i 213. 200. 87. 254 0 3257 701 80 i 205. 215. 45. 50 0 4006 6461 3786 i 195. 66. 225. 254 0 5459 6461 3786 i 203. 62. 248. 4 0 1221 3786 i 167. 142. 3. 6 0 5056 6461 3786 i 195. 219. 96. 239 0 8297 6461 3786 i 195. 211. 29. 254 0 5409 6461 3786 i AS 80 is General Electric, AS 701 is UUNET, AS 7018 is AT&T AS 3786 is DACOM (Korea), AS 1221 is Telstra

Characteristics of the AS Graph • AS graph structure – High variability in node degree (“power law”) – A few very highly-connected ASes – Many ASes have only a few connections CCDF 1 All ASes have degree >= 1 0. 01 Very few have degree >= 100 0. 001 1 10 1000 AS degree

Characteristics of AS Paths • AS path may be longer than shortest AS path • Router path may be longer than shortest path 2 AS hops, 8 router hops d s 3 AS hops, 7 router hops

Problem of Missing Edges • Limited collection of paths – Some edges might never be traversed – Especially low in the AS hierarchy – … and backup links • Example: paths from two tier-1 ISPs miss an edge Sprint AT&T d 1 Harvard ? ? ? Harvard B-school d 2

Problem of Missing Nodes • Route aggregation – AS advertises a larger address block – Smaller address block not seen everywhere – Can cause an AS not to appear in BGP table C: 12. 0. 0. 0/8 A B D D: 12. 1. 0. 0/16 C E E: 12. 2. 0. 0/16 – C’s table: has paths “C”, “C D”, and “C E” – B’s table: has only path “C” for 12. 0. 0. 0/8

Research Questions • Incomplete data – How to get more data? – How much does missing data affect answers? – What kinds of questions can be answered safely?

AS Relationships

Interdomain Routing Policies • Two main decisions – Path selection: which of the paths to use? – Path export: which neighbors to tell? • Both driven by business relationships, e. g. , – Customer pays provider for Internet access – Peers find it mutually advantageous to cooperate “ 12. 34. 158. 0/24: path (2, 1)” 3 “ 12. 34. 158. 0/24: path (1)” 1 2 data traffic 12. 34. 158. 5

Customer-Provider Relationship • Customer needs to be reachable from everyone – Provider exports routes learned from customer to everyone • Customer does not want to provide transit service – Customer does not export from one provider to another Traffic to the customer Traffic from the customer d provider advertisements provider traffic customer d customer

Peer-Peer Relationship • Peers exchange traffic between customers – AS exports only customer routes to a peer – AS exports a peer’s routes only to its customers Traffic to/from the peer and its customers advertisements peer d traffic peer

Paths You Should Never See (“Invalid”) Customer-provider Peer-peer two peer edges transit through a customer

Other Kinds of Relationships • Siblings – – Same company Mutual transit service Like one bigger AS Mergers, acquisitions, … • Backup – Used only when failure – Second provider – Backup peering • Geography-specific – Customer in U. S. – Peer in Europe E A B H C D F G primary backup

AS Relationships Matter • Scientific understanding – Understanding Internet structure and evolution – Understanding why certain paths are used for traffic • Placement of Web servers – Want to be close to most customer networks • Business decisions – Selecting new peer or provider, or renegotiating relations • Security policies – Knowing which BGP routes look suspicious • Analyzing BGP convergence – Relationships have a big impact here (more later!)

Inferring AS Relationships • Top down: how routes are selected – AS relationships define routing policy – Routing policy determines the routes you see • Bottom up: how policies can be inferred – Routing data are available from public sources – The chosen routes tell you about the policy • Example: seeing path “A B C” tells you… – B permits A to transit through B to reach C – (A, B) and (B, C) cannot both be peering links – A and C are not both upstream providers of B

Type-of-Relationship Problem • Given the inputs – AS graph G(V, E) with vertices V and edges E – Set of paths P on the graph G • Find a solution that – Labels each edge with an AS relationship – Minimizes the number of “invalid” paths in P • Rich area of research work – http: //www-unix. ecs. umass. edu/~lgao/ton. ps – http: //www. cs. princeton. edu/~jrex/papers/infocom 02. pdf – http: //www. cs. berkeley. edu/~sagarwal/research/BGP-hierarchy/ – … lots of scope for algorithms-oriented research project

AS Relationship Graph (2002) • Lowest level: Stubs – Leaf nodes: no peers or downstream customers – 8898 of the 10915 ASes (82. 5% of ASes) – Ex: UC Berkeley (25), Princeton (88), AT&T Labs (6431), and INRIA (1300) • Next lowest level: Regional ISPs – Leaf nodes after successive pruning of leaf nodes – 971 ASes of the 10915 ASes (8. 9% of ASes) – Ex: Pac. Bell (5676), US West (6223), and UUNET Canada (815) • Remaining 1046 ASes: Core

AS Relationship Graph (2002), Continued • Dense core: Tier-1 providers – (Nearly) fully-connected nodes with no providers – Around 15 -20 ASes in a near-clique – Ex: Sprint, UUNET, AT&T, Verio, Level 3, C&W, … • Transit core: – ASes that peer with the dense core and each other – 129 ASes, including top providers in Europe and Asia – Ex: UUNET Europe, KDDI, and Singapore Telecom • Outer core – All of the remaining ASes in the core – 897 ASes, including large regional and national ISPs – Ex: Turkish Telecom and Minnesota Regional Network

Node Degree is Not Enough • Node degree ignores relationships – A stub AS may have many upstream providers – A core AS may have a small number of peers – Some ASes have customers that don’t have AS numbers

Ideas for Class Projects • AS relationship inference – New algorithms for inferring AS relationships – Longitudinal study of AS relationship graph – Influence of policies on measuring AS graph • AS peering policies – Analysis of incentives to peer or not – Study of how one AS can game another – Analysis of whether regulation is necessary to keep large ASes from locking out smaller ones • Alternate settlement models – Associating prices with routes? – Source-based routing with third-party control?

Peering Policies • Contracts that outline: – Operational requirements on peer network – Backbone and peering capacity requirements – Number and geographic diversity of peering points – Volume and ratio of traffic between two peers – Routing-policy requirements • AOL’s Settlement-Free Interconnection Policy – http: //www. atdn. net/settlement_free_int. shtml

AOL’s Settlement-Free Interconnection Policy • Operational requirements on a peer network – Handle a single-node outage w/o traffic impact – Single AS number – Network Operations Center staffed at all times • Backbone capacity – At least 10 gigabits/sec between 8 or more cities – Minimum peering link speed of 622 megabits/sec • Peering locations (in U. S. ) – At least four locations – Must include D. C. area, middle of country, Bay area, and NYC or Atlanta

Efficient Early-Exit Routing • Diverse peering locations Customer B – Both costs, and middle • Comparable capacity at all peering points Provider B – Can handle even load multiple peering points • Consistent routes Early-exit routing Provider A Customer A – Same destinations advertised at all points – Same AS path length for a destination at all points

AOL Routing Requirements • Consistent advertisements – All customer routes – At all peering points – With the same AS path length • Address blocks – Routes aggregated as much as possible – No address blocks smaller than /24 – Address blocks are registered (e. g. , with ARIN) • No default routing – Only send traffic to destinations AOL advertises

For Next Tuesday’s Class • Adapting routing inside an AS to the traffic – “The Revised ARPANET Routing Metric” (1989) – “Traffic Engineering With Traditional IP Routing Protocols” (survey paper, 2002) • Written one-page review of first paper – Brief summary of the paper – Reasons to accept the paper – Reasons to reject the paper – Suggestions for future research directions • Just to skim… – RFC 3272: “Overview and Principles of Internet Traffic Engineering”