Advanced Networking Lab Given two IP addresses the estimation algorithm for the path and latency between them is as follows Step 1 Map IP addresses to AS numbers We use BGP routing tables to map an IP address to an AS number Step 2 Infer AS paths between ID: 509091
Download Presentation The PPT/PDF document "Path Stitching: Internet-Wide Path and D..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Path Stitching: Internet-Wide Path and Delay Estimation from Existing Measurements
Advanced Networking Lab.
Given two IP addresses, the estimation algorithm for the path and latency between them is as follows: Step 1: Map IP addresses to AS numbers. We use BGP routing tables to map an IP address to an AS number. Step 2: Infer AS paths between
Ases
. We use Qiu and Gao's methodology to infer AS paths between two ASes. Their methodology exploits the AS paths appeared in BGP routing tables. Step 3: Stitch path segments along the inferred AS path. Taking as input the inferred AS path from Step 2, we stitch the individual path segments as returned from the index. Here, the ending and beginning points of the segments must be exact IP address matches. We discuss approximation methodologies for cases without exact matches . Step 4: Rank stitched paths and select the best. Out of the multiple stitched paths from Step 3, we select the closest path to the real path.
Path stitching
Web interface
Demo scenario
Massively distributed applications have become very popular in today's Internet. These applications require constantly updated information about the network-internal performance characteristics, such as round-trip delay, effective bandwidth, IP hop count, and loss rate. Today's Internet does not provide any such information, and applications and new services resort to proprietary measurements to obtain necessary information. Despite their novel methodologies of Internet-wide measurement and network performance estimation, these measurement systems are usually designed without much consideration for compatibility or inter-operability. Instead of performing measurements ourselves and generating active probes, we do no additional measurements on our own (other than for evaluation). We utilize existing measurements and extrapolate estimates for uncharted points.
Our
work is orthogonal to existing measurement systems, as our goal is to integrate diverse data sets and others' is to improve the accuracy in measurement. In this
work
we propose an algorithm to combine data sets and produce estimates. We focus only on delay and path in this work. Two data sets we consider are: hop-by-hop delay measurements from traceroute and BGP (Border Gateway Protocol) routing tables. The main idea is to segment traceroute outputs by the AS, and stitch up a path for a query using BGP routing tables. Fittingly, we call our algorithm path stitching. In every step of our algorithm, we face many sources of error. We illustrate challenges in dealing with those errors, and present a few heuristics to address them. Despite many sources of error, the evaluation results are promising.
Abstract
Router,
PoP, or IP Prefix Clustering
Intra-domain segment
Inter-domain segment
stitched path
End-to-end delay graphs are generated on-the-fly
Display statistics of
stitched
paths
Stitched paths are displayed and animated
on Google Map
Query interface
BGP routing table
snapshots
Traceroute
outputs
(IPv4 Routed /
24 topology
data
set)
toronto.kaist
sibilla
daemon
prototype
RouteViews and RIPE
CAIDA Ark project
Contributors
Internet Sibilla
demo-site.snu
Internet Users
Web browser
(http://cannes.kaist.ac.kr/sibilla)
Data contribution
Query
Reply
Internet
Sibilla
*: Utilizing DNS for Delay Estimation Service
In this work, we propose to design an end-to-end delay estimation service, called
Sibilla
. The novelty of our system is that we utilize existing measurements and an infrastructure. Our delay estimation is based on a simple idea of path stitching. To provide the path-stitching as a service, we design and deploy our estimation service system on top of the Domain Name System (DNS). We deploy a
Sibilla
DNS server that receives a query between two hosts and replies with an estimated delay through the DNS interface. We take advantage of the fact that local DNS cache servers are distributed globally and exist in almost all
ASes
and improve the accuracy of our path stitching algorithm.
Abstract
*
Sibilla
is a female oracle figures in Greek and Roman times, offering cryptic predictions to those who sought to know the future (Sybil in English)
The basic
idea of our service system is to formulate a
delay query
between two hosts as a DNS query and let our Sibilla DNS server answer it. For example, when a user wants to estimate delay between two points A and B, the user submits a DNS query for A_B.latency.sibilla.com. The local DNS server forwards the query to our Sibilla DNS server, which in turn replies with delay estimation.
Sibilla
overview