We have proposed techniques for improving how IP addresses of network infrastructure are mapped to the administering Autonomous Systems
Towards an accurate AS-level traceroute tool
SIGCOMM, no. 4 (2003): 365-378
Traceroute is widely used to detect routing problems, characterize end-to-end paths, and discover the Internet topology. Providing an accurate list of the Autonomous Systems (ASes) along the forwarding path would make traceroute even more valuable to researchers and network operators. However, conventional approaches to mapping traceroute...更多
下载 PDF 全文
- Network operators and researchers would benefit greatly from an accurate tool for reporting the sequence of Autonomous Systems (ASes) along the path to a destination host.
- Designing a useful “ASlevel traceroute” tool depends on having an accurate way to map the IP addresses of network equipment to the administering ASes. Designing a useful “ASlevel traceroute” tool depends on having an accurate way to map the IP addresses of network equipment to the administering ASes
- This problem is surprisingly difficult and existing approaches have major limitations, due to the operational realities of today’s Internet.
- This improved IP-to-AS mapping can be used as seed input for a tool that maps traceroute output to an AS-level path
- Network operators and researchers would benefit greatly from an accurate tool for reporting the sequence of Autonomous Systems (ASes) along the path to a destination host
- Improved IP-to-AS mapping: Many mismatches between BGP and traceroute paths can be explained by Internet eXchange Points, sibling Autonomous Systems managed by the same institution, and Autonomous Systems that do not advertise routes to their equipment
- We have proposed techniques for improving how IP addresses of network infrastructure are mapped to the administering Autonomous Systems
- These techniques rely on a measurement methodology for (i) collecting both BGP and traceroute paths at multiple vantage points and using an initial IP-to-AS mapping derived from a large collection of BGP routing tables
- We presented heuristics that compare the BGP and traceroute AS paths to identify Internet eXchange Points, sibling Autonomous Systems, and other Autonomous Systems that “share” address space, and evaluated the improved IP-to-AS mapping on traceroute paths collected from three vantage points
- We were able to include more than 99% of the BGP AS paths in our analysis because most BGP routes are relatively stable and few BGP AS paths have private Autonomous Systems or AS SETs
- Compared to an initial IP-to-AS mapping constructed from the BGP tables, our heuristics reduced the fraction of incomplete paths from 18–22% to 6–8%; the ratio of matched to mismatched paths more than doubled, increasing from around 9–12 to 25–35
- The authors present the methodology for collecting traceroute and BGP paths from multiple vantage points, as shown in Figure 1.
- Starting with a list of routing table entries, the authors first identify the prefixes that cover the routable address space and select two IP addresses within each prefix for traceroute probing.
- A table with routes for 18.104.22.168/8 and 22.214.171.124/10 would only use the 126.96.36.199/8 routing entry for destinations in 188.8.131.52/9 or 184.108.40.206/10; all other addresses in 220.127.116.11/8 would match 18.104.22.168/10
- To identify these cases, the authors sort the list of prefixes based on the numerical values and mask length; this ensures that each prefix is followed immediately by all of its subnets.
- Less than 8.3% of the traceroute AS paths differ from the corresponding BGP AS path.
- After applying the new IP-to-AS mapping to all of the traceroute paths, 85.9-87.0% of the traceroute AS paths matched the corresponding BGP AS paths.
- This increase came from up to a 12% reduction in the mismatched paths and up to a 50% reduction in the incomplete paths.
- The authors were able to include more than 99% of the BGP AS paths in the analysis because most BGP routes are relatively stable and few BGP AS paths have private ASes or AS SETs
- The authors have proposed techniques for improving how IP addresses of network infrastructure are mapped to the administering ASes. In this paper, the authors have proposed techniques for improving how IP addresses of network infrastructure are mapped to the administering ASes
- These techniques rely on a measurement methodology for (i) collecting both BGP and traceroute paths at multiple vantage points and using an initial IP-to-AS mapping derived from a large collection of BGP routing tables.
- The authors presented heuristics that compare the BGP and traceroute AS paths to identify IXPs, sibling ASes, and other ASes that “share” address space, and evaluated the improved IP-to-AS mapping on traceroute paths collected from three vantage points.
- The improved mapping helps in highlighting the small number of important cases when the traceroute and BGP AS paths differ
- Table1: Traceroute probing locations
- Table2: Number of prefixes in the three datasets
- Table3: Prefixes excluded due to BGP properties
- Table4: Ending of the traceroute experiments
- Table5: BGP vs. traceroute AS paths for different AS mapping techniques
- Table6: BGP tables for IP-to-AS mapping around May 2003 an address that did not match any prefix in the set of BGP tables. Private IP addresses accounted for less than 40% of the cases. Unmapped hops can arise when interfaces are assigned addresses that are not advertised to the larger Internet. “*” hop: Many traceroute paths had one or more “*” characters, even after removing trailing “*” characters at the end of the path. A “*” hop may stem from a lost probe or ICMP packet, or from an intermediate node that does not participate in ICMP. Multiple origin AS hop: Around 9–13% of the traceroute paths had at least one interface address that mapped to multiple AS numbers, making direct comparisons with the BGP path impossible. MOAS prefixes occur for various reasons including misconfiguration, multihoming, or exchange points [<a class="ref-link" id="c16" href="#r16">16</a>]. The three cases are not mutually exclusive; a single traceroute path may have hops with one or more of these properties
- Table7: Statistics on mismatched traceroute paths
- Table8: Patterns and possible causes of mismatched AS paths and AS H for the BGP path in Figure 2(d)
- Table9: AS numbers and prefixes inferred as IXPs isfy our criteria of an IXP; these cases are listed in Table 9 in decreasing order of fan-in and fan-out. To verify our results, we first queried whois using the AS number or prefix to see if the description contained the words “exchange point” or “Internet exchange”; for example, AS 5459 was listed as “London Internet Exchange” in whois.ripe.net. This check succeeded for 18 of our 25 inferences. Then, we compared our results against a list of known IXPs [<a class="ref-link" id="c25" href="#r25">25</a>]. This confirmed 16 of the 25 inferences. Together, 19 of the 25 inferences passed at least one of these checks. Some of the remaining cases (highlighted in italics) may be IXPs, too; for example, CalRen is an exchange point for universities in California
- Table10: The results of using the three techniques to tune the IP-to-AS mapping
- Table11: The effect of multiple vantage points: comparing using the first three with all eight probing locations
- Table12: Remaining mismatches with BGP AS path on the diversity of AS-level paths. Previous work [<a class="ref-link" id="c26" href="#r26">26</a>] studied the marginal utility of discovering network topology using traceroute. They concluded that increasing the number of sources in traceroute experiments has low utility beyond the second source. Increasing the number of sources is admittedly more important for our purposes, though, since our heuristics rely on fan-in as well as fan-out counts
- Recent measurement studies have quantified the differences between BGP and traceroute AS paths. The analysis in  showed that these differences have a significant impact on the characterization of the Internet topology. In parallel with our paper, the work in  used publicly-available data (such as whois, lists of known IXPs, and other Web sites) to test the hypothesis that many of the mismatches stem from IXPs and siblings; in contrast, our paper proposes heuristics for identifying IXPs, siblings, and other causes of mismatches to improve the IP-to-AS mapping. To improve the accuracy of AS graphs derived from traceroute, the work in  proposed techniques that identify border routers between ASes to correct mistaken AS mappings; this is an alternate approach that handles some of the inaccuracy introduced by IP-to-AS mappings derived from BGP tables. Traceroute data have been used in other studies that measure router-level topologies and map routers to ASes [4, 6]. Except for handling certain traceroute anomalies such as unmapped IP address, these studies did not focus on improving the accuracy of the IP-to-AS mapping derived from the BGP routing tables. Focusing solely on BGP AS paths, the work in [19, 20] presented algorithms for inferring AS-level commercial relationships, including siblings; however, these studies did not consider the influence of sibling ASes on the accuracy of traceroute AS paths.
- Van Jacobson, “Traceroute,” ftp://ftp.ee.lbl.gov/traceroute.tar.gz.
- V. Paxson, “End-to-End Routing Behavior in the Internet,” IEEE/ACM Trans. Networking, vol. 5, no. 5, pp. 601–615, October 1997.
- Stefan Savage, Andy Collins, Eric Hoffman, John Snell, and Tom Anderson, “The end-to-end effects of Internet path selection,” in Proc. ACM SIGCOMM, September 1999.
- Ramesh Govindan and Hongsuda Tangmunaraunkit, “Heuristics for Internet map discovery,” in Proc. IEEE INFOCOM, 2000.
- “Skitter,” http://www.caida.org/tools/measurement/skitter.
- Neil Spring, Ratul Mahajan, and David Wetherall, “Measuring ISP topologies with Rocketfuel,” in Proc. ACM SIGCOMM, August 2002.
- Paul Barford, Azer Bestavros, John Byers, and Mark Crovella, “On the marginal utility of network topology measurements,” in Proc. Internet Measurement Workshop, November 2001.
- Ratul Mahajan, David Wetherall, and Tom Anderson, “Understanding BGP misconfigurations,” in Proc. ACM SIGCOMM, August 2002.
- “Visualizing Internet topology at a macroscopic scale,” http://www.caida.org/analysis/topology/as_core_network/.
- Hongsuda Tangmunarunkit, Ramesh Govindan, Scott Shenker, and Deborah Estrin, “The impact of policy on Internet paths,” in Proc. IEEE INFOCOM, 2001.
- Lisa Amini, Anees Shaikh, and Henning Schulzrinne, “Issues with inferring Internet topological attributes,” in Proceedings of SPIE, July 2002, vol. 4865.
- Timothy G. Griffin and Gordon Wilfong, “On the correctness of iBGP configuration,” in Proc. ACM SIGCOMM, August 2002.
-  “Prtraceroute,” http://www.isi.edu/ra/RAToolSet/prtraceroute.html.
-  Paul Barford and Winfred Byrd, “Interdomain routing dynamics,” Unpublished report, June 2001.
-  Xiaoliang Zhao, Dan Pei, Lan Wang, Dan Massey, Allison Mankin, S. Felix Wu, and Lixia Zhang, “An analysis of BGP multiple origin AS (MOAS) conflicts,” in Proc. Internet Measurement Workshop, November 2001.
-  Young Hyum, Andre Broido, and k claffy, “Traceroute and BGP AS Path incongruities,” 2003. http://www.caida.org/outreach/papers/2003/ASP/.
-  Hyunseok Chang, Sugih Jamin, and Walter Willinger, “Inferring AS-level internet topology from router-level path traces,” in Proc. Workshop on Scalability and Traffic Control in IP Networks, SPIE ITCOM Conference, August 2001.
-  L. Gao, “On inferring autonomous system relationships in the Internet,” IEEE/ACM Trans. Networking, December 2001.
-  Lakshminarayanan Subramanian, Sharad Agarwal, Jennifer Rexford, and Randy H. Katz, “Characterizing the Internet hierarchy from multiple vantage points,” in Proc. IEEE INFOCOM, June 2002.
-  C. Labovitz, A. Ahuja, A. Bose, and F. Jahanian, “Delayed Internet routing convergence,” IEEE/ACM Trans. Networking, vol. 9, no. 3, pp. 293–306, June 2001.
-  “University of Oregon Route Views Project,” http://www.routeviews.org/.
-  “Ripe NCC,” http://www.ripe.net/ripencc/pub-services/np/ris/.
-  “Singaren,” http://noc.singaren.net.sg/netstats/routes/.
-  “Packet Clearing House,” http://www.pch.net/resources/data/exchange-points/.
-  Paul Barford, Azer Bestavros, John Byers, and Mark Crovella, “On the Marginal Utility of Network Topology Measurements,” in Proc. Internet Measurement Workshop, November 2001.
-  E. Chen and J. Stewart, “A Framework for Inter-Domain Route Aggregation,” Request for Comments 2519, February 1999.
-  J. Postel, “Internet Control Message Protocol,” RFC 792, September 1981.