Projects of Internet Measurement

CAIDA: The Cooperative Association for Internet Data Analysis
DIMES
Chris Harrison's Internet Map
IPv4 Address Report
BGP Report
iPlane: An Information Plane for Distributed Services
Rocketfuel: An ISP Topology Mapping Engine
INTEL/DANTE (Delivery of Advanced Network Technology to Europe)
Sarangworld Traceroute Project Pattern Matching
Route Views

Some method to measure:
latency: RTT
pkt loss: active probing
IP geo-location: whois, coutry, city, and ZIP code(CAIDA NetGeo)..
link bandwidth: pkt dispersion (equal size pkt pair/train), P2P peers' uploading rate to me(iPlane)..
link capacity: varying the size of pkt pair (pathchar).
active probing pkt: ICMP-more dests, UDP-more links, least dests(CAIDA's Ark@CATCH).

Some paper notes:
1. CAIDA's Ark@CATCH, Internet Mapping: from Art to Science
-31 monitors, LINDA-like coordination structure
-mainly traceroute to every /24 to get router-level topology
-UDP probing for router alias resolution
-then map to AS-level using BGP info from Routeview
-rDNS to get geo-info and etc.

2. CAIDA@SIGCOMM'05 workshop: Classifying the Types of Autonomous Systems in the Internet
Companies: 41.6%
ISP: 34.33%
Univ.: 10.49%
NIC(Network Information Center): 6.86%

3. CAIDA@SIGCOMM'07 CCR: AS Relationships: Inference and Validation
c2p(customer to provider): 3070
p2p(peer to peer): 623
s2s(sibling to sibling): 31

4. CAIDA: Measuring the use of IPv4 space with Heatmaps
data from 1hour two tier1 ISP backbone links:
-Chicago->Seattle
-Los Angeles->San jose
observation:
-large scanning/DOS/spoofing traffic
-legacy address space is not that well utilized (similar to our measurement!).

5. The "robust yet fragile" nature of the Internet@PNAS
SF model VS HOT model of Internet topology.
-SF (scale-free): power-law: k * d_k^a = c, a and c are constant
highly connected vertices are crucial to the global connectivity and they are where most traffic pass through.
-HOT (highly optimized/organized tolerance/tradeoffs)
maximize utility with limited resources
backbone cost is primarily dominated by installation of links
result is highly variable bandwidth and router degrees at network's periphery, but a much greater uniformity of high-bandwidth and low-degree routers in the core.
They claim that the real network is more like HOT but not SF.