Chair of Network Architectures and Services Department of Informatics Technical University of Munich
Advanced Computer Networking (ACN)
IN2097 – WiSe 2020-2021
Prof. Dr.-Ing. Georg Carle
Sebastian Gallenmüller, Max Helm,
Benedikt Jaeger, Patrick Sattler, Johannes Zirngibl
Chair of Network Architectures and Services
Department of Informatics
Technical University of Munich
Internet Structure & Measrdroid
Packet networks
Research on Internet structure
Example analysis: environmental influences on data quality
Bibliography
Internet Structure & Measrdroid
Packet networks
Research on Internet structure
Example analysis: environmental influences on data quality
Bibliography
Packet networks
Arpanet
Packet networks
Arpanet 1969
•
Interface Message Processors (IMPs) as
gate-ways,
performing
store-and-forward
packet
switching
•
IMPs from BBN, based on Honeywell 316
mini-computers, 16-bit CPU, 24 Kbyte memory
•
Host computers connected to IMPs via serial
in-terfaces
•
Interconnection of IMPs with modems connected
to leased lines, 50 kbit/s
•
Computers attached: SDS Sigma 7, SDS 940,
DEC PDP-10, IBM 360/75
Packet networks
Arpanet 1977
Packet networks
MILNET 1989
Packet networks
Internet today
Packet networks
Internet Components
•
Billions of connected computing devices
• hosts=end systems • runningnetwork apps
•
Communication links
• fiber, copper, radio, satellite • transmission rate =bandwidth
•
Routers:
• forward packets (chunks of data)
Packet networks
Internet Dimensions
•
Protocols
control sending, receiving of messages
• e.g., Ethernet, IP, TCP, HTTP, Skype
•
Internet: "network of networks"
• loosely hierarchical
• public Internet versus private Internet
•
Internet standards
• RFC: Request for comments • IETF: Internet Engineering Task Force • IANA: Internet Assigned Numbers Authority
•
Communication infrastructure
enables distributed applications:
• Web, VoIP, e-mail, games, e-commerce
•
Communication services provided:
• reliable data delivery from source to destination
• "best effort" (unreliable) data delivery
Packet networks
Acknowledgements
•
Jim Kurose, University of Massachusetts, Amherst
•
Keith Ross, Polytechnic Institute of NYC
•
Olivier Bonaventure, University of Liege
•
Srinivasan Keshav, University of Cambridge
Internet Structure & Measrdroid
Packet networks
Research on Internet structure
Example analysis: environmental influences on data quality
Bibliography
Research on Internet structure
Why measrdroid?
•
How to learn about the Internet topology
X Passive evaluation of BGP tables X Evaluation of meta data (e.g. RIPE DB) I Distributed traceroute measurements•
Reasonable Approach, measure the Internet from as many
vantage points as possible:
X Public looking glasses X PlanetLab X Rented servers
X Operating an autonomous system (AS) at Internet exchange points
I Mobile measurement framework
Research on Internet structure
What is measrdroid?
•
MeasrDroid
• Free measurement client with focus on networks
• Utilization of integrated sensors and API functions • Incorporation of native C code where beneficial
• Designed for interdisciplinary research
• Project lead: Johann Schlamp ([email protected])
•
Progress so far:
• >10 students worked in the project,
advised by Johann Schlamp (Bachelor thesis / Master thesis / research assistant)
• >32,000 Lines of Code (in 35 packages)
• Periodic measurements of 335 distinct data points:
hardware-environment-location-network-telephony
Research on Internet structure
MeasrDroid
•
Sensors for measurement of...
• network (configuration, wifi, stats, NAT testing, active
mea-surements)
• device info (model, device, version, hardware, brand, battery,
nfc . . . )
• telephony (cell data, sim data, neighbour cells) • location (current location, gps status)
• environment (accelaration, temperature, gyroscope, gravity,
pressure, proximity, humidity, magnetic field, light)
•
Analysis of data to...
• understand changes over time • find correlations in the data
Research on Internet structure
MeasrDroid data collection
•
The measrdroid app collects approximately 335 single data points per measurement (http://www.droid.net.in.tum.de/about/privacy)
• Hardware Info:
Information about the device hardware, e.g., model, manufacturer, CPU, RAM . . .
• Environment Sensors:
Environment sensors provided by hardware (some might not be installed)
• Location:
Details about current location and positioning techniques
• Network Info:
Technical data about network connection, including active measurements
• Device Status:
The current status of your device, e.g. battery status
• Task details:
Details about a specific measurement task provided by our researchers
• Telephony:
Various data about your current signal strength
• MeasrDroid usage & core settings:
Application-specific usage data and your measurement preferences
Research on Internet structure
MeasrDroid website
http://droid.net.in.tum.de (Sept. 2016)
Research on Internet structure
MeasrDroid map (Europe)
Research on Internet structure
MeasrDroid
•
Alpha phase in ’12
•
Several Beta tests from
May ’13 – January ’14
•
Google Play release in March ’14
•
>200 participating clients
•
1,220,000 data sets
• Clients traveled nearly a million kilometers
• At an average speed of >70 km/h
•
Discovery of 18,612 CIDs
Research on Internet structure
MeasrDroid SSIDs
Research on Internet structure
MeasrDroid map Germany
Research on Internet structure
MeasrDroid map Munich
Research on Internet structure
AS graphs
Research on Internet structure
CDN analysis
•
How is Google connected to the (wired) Internet?
• Mean AS path length: 2-3 • Mean IP path length: 12
•
Is there a difference to mobile connections?
• Mean AS path length: 1-2 • Mean IP path length: 14
•
What about DNS resolution for google.com?
• 566 IP addresses spread over two /16 blocks (observed in 2 months of measurement)
• Wired networks: 250 distinct IP addresses
• Mobile networks: 500 distinct IP addresses
•
What about RTTs to Google?
Michael Faath.„Analysis of content delivery networks with an Android-based measurement framework“.Magisterarb. Germany: Technical University of Munich, 2013
Research on Internet structure
Example Google
•
Let’s have a look at DNS
• What IP address corresponds to google.de? • Answer: 566 IP addresses spread over two /16 blocks
(observed in 2 months of measurement)
•
Again: Is there a difference between wired and mobile connection?
• Wired networks: 250 IP addresses observed • Mobile networks: 500 IP addresses observed
•
Maybe there is even a provider-specific difference?
• O2 network: 328 IP addresses observed • Simyo network: 323 IP addresses observed
• T-Mobile network: 162 IP addresses observed
Research on Internet structure
Example Google
Round trip time (ping) measurements to
google.de
over a period of two days
•
Better RTTs for newer radio technologies
•
Not really surprising
Internet Structure & Measrdroid
Packet networks
Research on Internet structure
Example analysis: environmental influences on data quality
Bibliography
Example analysis: environmental influences on data quality
Localization accuracy
•
Comparison of
phone-reported accuracy
vs.
actual deviation
in a controlled experiment
•
Reason for spikes?
Example analysis: environmental influences on data quality
Localization accuracy
Example analysis: environmental influences on data quality
Localization accuracy
Example analysis: environmental influences on data quality
Localization accuracy
Example analysis: environmental influences on data quality
Localization accuracy
Example analysis: environmental influences on data quality
Signal quality - handover influence
Fluctuating signal strength on fast train (RSSI in range [0, 32])
Many handovers even at fixed positions
Example analysis: environmental influences on data quality
WiFi AP localization
•
Can we reverse-engingeer Google’s Wifi localization technique?
• Try to triangulate Wifi access points (APs) • Map signal strengths to radii and intersect circles
• Compare results to publicly available ground truth (open wifi spots), evaluate Telekom WLAN-to-go?
Example analysis: environmental influences on data quality
GSM cell tower localization
•
Utilize Cell IDs and measured signal strengths
• Try to identify coverage area for single antennas
• Estimate direction of signal deterioration • Find source of signal
• Map multiple cells to a single tower