Visualizing Network Data

1
Visualizing Network Data

• R. A. Becker, S.G. Eick, A.R. Wilks

Reviewed by Bill Kules and Nada Golmie for CMSC
838S Fall 1999
2
Outline

• What is a network?
• Challenges in visualizing large networks
• Early work knowledge bases (Fairchild)
• Motivation for SeeNet
• SeeNet project description
• Critique
• State of the Art
• Demos

3
What is a Network?

• Communication networks
• Internet, telephone network, wireless network.
• Network applications
• The World Wide Web, Email interactions
• Transportation network/ Road maps
• Relationships between objects in a data base
• Function/module dependency graphs
• knowledge bases

4
Challenges in Visualizing Large Networks

• Positioning nodes
• Managing link/ information
• Graph scales
• Navigation/ interaction

Layout of Internet routes and IP addresses from
data collected in September 1998, appeared in
Wired Magazine December 1998 issue
5
Early Work - Fairchild (1988)

• Representation of knowledge bases
• relationships among objects are represented as
  directed graphs in 3D space.
• Platform requirements
• identification of individual elements
• relative position of an element within a context
• explicit relationships between elements
• Main issues investigated are
• Positioning, coping with large bases (Fisheye
  views), navigation and browsing, dynamic
  execution of knowledge base.

6
Motivation for SeeNet

• SeeNet is a monitoring and visualization tool to
  display and analyze large volumes of network data
  and statistics (ATT long distance network
  traffic).
• Overcome the display clutter problem associated
  with large networks
• Interactive techniques
• More traditional methods such as aggregation,
  averaging, and thresholding.

7
SeeNet Project Description

• SeeNet is designed to address the display clutter
  problem. It consists of a collection of graphical
  tools that include techniques for
• Static Display
• Interactive Controls
• Animation

8
Static Network Display Features

• Linkmaps
• too complex resulting in display clutter problem
• Nodemaps (glyphs)
• node contains information/statistics
• tradeoffs with details information about
  particular links
• Matrix Display
• to/from nodes are assigned row/columns and matrix
  cells are associated with links.
• Solves visual prominence and overplotting problem
• Gives up geographical information
• ordering of rows/columns may be important

9
Parameter Focusing

• Controls network display characteristics and
  provides dynamic parameter adjustment. Parameter
  values and classes include
• statistic, levels (probing, brushing)
• geography/ topology (zoom)
• time (averaging), aggregation, size and color.
• Main problems are
• large range of values
• multi-parameters lead to confusing displays
• displays are sensitive to particular parameter
  values

10
Direct Manipulation

• Modify focusing parameters while continuously
  provide visual feedback and update display (fast
  computer response).
• Features include
• Identification highlight, color, shape
• Linkmap parameter control line thickness,
  length, color legend, time slider, animation

11
More Direct Manipulation Features

• Matrix display parameter control drag-and-drop
  action, row/column reordering.
• Nodemap parameter control symbol size, color
• Animation analysis of time-varying data
• Zooming and Birds Eye view
• Conditioning filtering, setting background
  variables and displaying foreground parameters.
• Sound

12
Application Examples
Worldwide Internet Traffic. Traffic on the
Internet, square root of packets transmitted from
country to country across the NSFNET backbone
during the first week of February 1993.
Department Email Communication Patterns. Each
node corresponds to a user, and links encode the
number of electronic mail messages sent between
the users.
13
Favorite Sentence

• Our goal is to understand the data and not the
  networks themselves.

14
Strengths

• Integrated techniques in one tool
• Node placement
• Graph scaling
• Manipulation
• Reducing clutter
• Good use of color
• Good overview of related work paper presentation
  is clear and well positioned in context.

15
Weaknesses

• Model is tailored for data networks
• Limited positioning capabilities
• New Jersey
• Doesnt work as well for e-mail example
• Novice vs. power user
• Could be tedious to adjust parameters
• need to use it to find out

16
Discussion

• Are there better ways of achieving our objective?
• Did we already know what we just learned?
• What is the predictive or insight value?
• Who would use the tool? Network engineer?
• Allows identification of interesting parameters,
  but could be limited beyond that.
• Paradox of graph visualizations
• Adjacent nodes are seen as more related, but long
  links are visually dominant
• How to explore structure - Fairchild
• How to explore statistics - Becker

17
State of the Art Network Visualization Tools

• Network management
• Traffic management and monitoring
• Internet statistics, traffic analysis
• Performance measurement
• Application /program performance (e.g. TCP/IP)
• Graph editing
• Relational databases
• A list of tools is available from CAIDA (SDSC)
  http//www.caida.org/Tools

18
Demos

• Internet Map
• Visual Route
• AS
• Skitter