Prior work has shown that bad things can happen (e.g., significant latency increases) when Internet paths change. When such path changes happen, which autonomous system (AS) you blame? Since BGP is an information-hiding protocol, finding the root cause of an interdomain path change is non-trivial. Prior work showed that the root cause might be any AS on the old or new paths. Thus, the potential set of root causes can be found by subtracting the intersection of the ASes on the two paths from the union of the ASes in the two paths.
However, this prior approach does not handle induced path changes. We need to revisit the assumption about the potential cause of path changes (being any AS on the old or new paths). The root cause may also be any AS on old paths from ASes in the new path from the vantage point or any AS on new paths from ASes in the old path from the vantage point. PoiRoot identifies alternative paths prior to a change occurring by using poisoning. An evaluation of PoiRoot on the BGP exchanges between five campuses showed a 100% accuracy and a precision (i.e., a mean root cause candidate set size) of 1.66.
Q: What do you see as the steps to deployment?
A: You need a prefix you can use for poisoning and vantage points.
Q: Is there any general purpose framework we can provide?
A: We use PlanetLab with reverse traceroute.
Q: Poisoning exposes operator's policies because it captures the alternative routes . Is this a problem?
A: Future work may be needed to support poisoning while preserving the privacy of operator policies. Nonetheless, PoiRoot may still beneficial in the presence of certain business relationships between ASes.
Q: When you bypass one AS you may also bypass its neighbors and then you cannot precisely determine the alternate paths.
A: We rely on what is happening in the unpoisoned case (determined using reverse traceroute) to determine how we should go about the poisoning process.
A: You need a prefix you can use for poisoning and vantage points.
Q: Is there any general purpose framework we can provide?
A: We use PlanetLab with reverse traceroute.
Q: Poisoning exposes operator's policies because it captures the alternative routes . Is this a problem?
A: Future work may be needed to support poisoning while preserving the privacy of operator policies. Nonetheless, PoiRoot may still beneficial in the presence of certain business relationships between ASes.
Q: When you bypass one AS you may also bypass its neighbors and then you cannot precisely determine the alternate paths.
A: We rely on what is happening in the unpoisoned case (determined using reverse traceroute) to determine how we should go about the poisoning process.
1z0-1081 Exam Dumps Simply make sure your grip on the IT braindumps devised the industry’s best IT professionals and get a 100% guaranteed success in Exam.
ReplyDelete