Doubletree: a Way to Reduce Redundancy in Traceroute Systems

Benoit Donnetjoint work with P. Raoult, T. Friedman and

M. Crovella

CAIDA – San Diego (Sept. 2005)

Context

● Network measurement● Internet topology discovery using distributed

traceroute monitors– IP interface level

● Existing tools:– Skitter (CAIDA)

– TTM (RIPE NCC)

– AMP (NLANR)

– DIMES (Tel Aviv U.)

Problem: Intra-Monitor Redundancy

Problem: Inter-Monitor Redundancy

Statement

● Classical approaches are out of date● Wide scale deployment of traceroute tools

– DIMES

● Scale up the number of monitors– little work on efficiency

● Be careful regarding– network resources

– load on destinations

Doubletree: Basics

● Cooperative algorithm● Goal: avoiding paths already explored● Exploits tree-like structure of routes in the

internet:– from a monitor to a set of destinations:

● monitor-rooted tree

– from a set of monitors to a destination: ● destination-rooted tree

Doubletree: Monitor-Rooted Tree

Doubletree: Destination-Rooted Tree

Doubletree: Reconciliation

● Backward and forward probing are opposite schemes

● How can we reconcile them?– Starts probing at some hop h

● First, forward probing from h● Second, backward probing from h-1

● How to choose h?– Probability p of hitting a destination with the very

first probe sent

Doubletree: Reconciliation (2)

Doubletree: Stop Sets

● Not necessary to maintain the whole tree structure.

● Each monitor uses stop sets: {(interface, root)}– Local Stop Set B: {interface}

● Backward probing

– Global Stop Set F: {(interface, destination)}● Forward probing● Shared between monitors● Could be implemented as Bloom filter

Doubletree: Coverage

Doubletree: Redundancy

Doubletree: Amount of Probes

Doubletree: Implementation

Future Works

● Improvement to Doubletree– Reducing the amount of probes sent

● Capturing the dynamicity of the network● BGP-guided probing● Overlay/P2P structure organizing the whole

system● Measurement tools standardization

Conclusion

● Doubletree is an efficient and cooperative algorithm allowing– redundancy reduction

– good coverage level

● Java implementation of a prototype