CSE561

1
CSE561 Graduate Computer Networking

• Venkat Padmanabhan
• Microsoft Research

2
Prelimaries

• Instructor Venkat Padmanabhan
• Email padmanab_at_cs.washington.edu
• Lectures MF, 1200-120 PM, EE1 003
• Office hours MF 130-200 PM, Sieg 226D
• TA Andrew Whitaker
• Email andrew_at_cs.washington.edu
• Office hours Th 230-330 PM, Sieg 433
• Home page http//www.cs.washington.edu/education/
  courses/561/01sp/

3
Whats this class about?

• Reading 25-30 papers spanning a wide spectrum of
  the networking research literature
• Learning how to critique networking research
• Learning how to do networking research through a
  hands-on project

4
Am I ready to take this class?

• If you have taken CSE 461 or an equivalent
  undergraduate class elsewhere, then youre all
  set.
• If you havent but have at least some background
  in computer systems, you can read through a basic
  textbook (e.g., the book by Peterson Davie)
  without much effort.
• Minimal coverage of basic material in class

5
Syllabus

• Focus on the Internet-related research
• Design philosophy and architecture
• Medium access control (Ethernet, wireless)
• Routing (BGP)
• Transport (TCP)
• Applications (Web/HTTP, audio/video)
• Novel paradigms (multicast, active networks)
• Novel networks (wireless, mobile)
• Security (protocol vulnerability, DoS attacks)

6
What you are expected to do

• Read and review research papers (20)
• Actively participate in discussions in class and
  on the mailing list (10)
• Take midterm exam (20)
• Define and work on a networking research project
  (50)

7
Literature Review

• 25-30 research papers
• all available online on the course home page
• only need to review papers marked as such
• optional reading need not be reviewed
• Review (one page max per paper)
• summarize paper in just a few sentences
• highlight contributions/lessons learned
• point out shortcomings in assumptions, approach,
  data, etc. (ignore presentation problems!)
• bonus points for comparing and contrasting with
  related literature (from reading list or
  elsewhere)

8
Literature Review (contd.)

• Reviews
• OK to discuss papers with others at any time
• But you must write your own review!
• Hardcopy due at the start of each lecture
• You may choose to post your review on the mailing
  list after the lecture period to get some
  discussion going

9
Class Participation

• Read the papers before coming to class
• Actively participate in discussion in class and
  on the mailing list
• Remember there are few absolute truths in
  networking. So dont be afraid to be a contrarian.

10
Mid-term Exam

• Scheduled for Friday, May 11
• Focus on understanding rather than the knowledge
  of facts

11
Project

• Define and work on a research project that
• is interesting
• explores novel issues or approaches
• is doable in the available time
• Groups of 2-3 students (not a rule)
• You are encouraged to come up with ideas
• feel free to brainstorm on ideas with others
• aim high! The best projects could lead to papers
  at the top conferences (SIGCOMM,Infocom,Mobicom)
• Ill distribute a list of potential projects in
  class next week

12
Project (contd.)

• Project scope
• implementation, simulation, measurement,
  analysis, or some combination thereof
• this is not about doing a literature survey
• The process
• initial meeting (April 6)
• 1-2 page project proposal due April 13
• problem statement, specific goals, resources you
  need, plan of attack, (anticipated) deliverables
• not a rigid document

13
Project (contd.)

• The process (contd.)
• checkpoint meetings after class on May 7
• project presentations and final report due the
  first week of June (exact date TBD)

14
Elements of a Network

• Links carry information (bits)
• Wire, wireless, fiber optic, smoke signals
• Switches move bits between links
• Routers, gateways, bridges, CATV headend, PABXs,
  
• Hosts are the communication endpoints
• PC, PDA, cell phone, toaster,

15
Example Local Area Network
Ethernet Hub

• Your home network
• Ethernet is a broadcast-capable multi-access LAN

Cable Modem
Printer
Laptop
PC
16
Example An Internetwork

• Internetwork is a network of networks
• The Internet is a global internetwork in which
  all participants speak a common language, IP.

ISP 2
ISP 1
Local Net 2
Local Net 1
17
Key Features of the Internet

• Distributed
• decentralized operation, global reach
• Large-scale
• 110 million hosts (Jan 2001) and growing
• Heterogeneous
• links, end-points
• Intelligent end-points
• supercomputers, PCs, PDAs
• Multi-purpose
• Web, audio/video conferencing, telnet

18
Growth of the Internet
19
Network Components

• Say you want download a file from host A onto
  host B.
• What problems do you need to solve?

20
Host Discovery

• Locating the remote host
• The naming problem
• provides a level of indirection between names and
  addresses
• Domain Name System (DNS)
• the telephone directory of the Internet

21
Medium Access Control

• Putting bits on the wire
• The network carries packets, not bits
• framing to create packets out of bits
• Physical medium may be shared
• example Ethernet, wireless
• medium access control (MAC) to arbitrate amongst
  several contenders
• Actually, the network only carries electrical or
  optical signals
• modulation to convert digital data into analog
  signals

22
Routing

• How do a packet find its way through the network?
• Internet is distributed
• no central brain
• routing happens in a decentralized manner
• Internet is not owned by any one organization
• hierarchical routing architecture
• intra-domain and inter-domain

23
Transport Protocol

• How do you ensure reliable transmission of the
  file over an unreliable network?
• Links and routers are a shared resource
• a packet may be dropped when the resource is
  over-utilized (or even otherwise!)
• resources need to be apportioned fairly
• Need to perform loss recovery and congestion
  control
• Transmission Control Protocol (TCP)
• responsible the vast majority of Internet traffic

24
Interaction of Transport Protocols with
Applications

• One size doesnt fit all
• Transfer size
• a protocol designed for large file transfers may
  be too heavyweight for exchanging short messages
• Reliability
• necessary for ordinary file transfer
• not necessary for live streaming audio

25
Multicast

• Several clients may want to download a file at
  the same time
• Repeated unicast would be inefficient
• Need to multicast the file
• avoid sending a packet over a link repeatedly
• Multicast makes existing problems harder,
  especially because of heterogeneity
• routing
• loss recovery
• congestion control

26
Resource Discovery

• Say you have a choice of where to download the
  file from
• Best to download it from nearest source
• this is exactly the goal of content distribution
  networks (e.g., Akamai)
• But Internet distance is not a static metric
  and is hard to measure

27
Security

• It aint a friendly world out there!
• Eavesdropping on conversations
• Masquerading as another entity
• man-in-the-middle attack
• Non-compliance with the protocols
• Deliberate overconsumption of resources
• denial of service attack

28
Mobility

• Hosts and/or networks may be mobile
• Examples
• a person carrying a laptop from work to home
• a planeload of people
• Maintaining Internet connectivity in the face of
  mobility
• mobility impacts most of the other components
  (discovery, routing, transport, applications,
  security)

29
Lecture Roadmap

• Internet design principles and architecture
• Medium Access Control
• Internet routing
• Router architectures
• Transport protocols congestion control
• Transport protocols interaction with
  applications
• Router support for congestion control
• Multicast routing

30
Lecture Roadmap (contd.)

• Multicast transport and applications
• Naming
• Resource discovery and selection
• Host and network mobility
• Wireless networks (medium access control)
• Wireless networks (TCP interactions)
• Network and protocol vulnerability
• Active networks

31
Topics Not Covered

• Quality of service
• Real-time communication
• Traffic measurement and analysis
• Ad-hoc networks
• and many others

32
Next Lecture Internet Design Principles and
Architecture

• History and evolution of the Internet
• Current architecture
• Defining principles
• layering
• datagram service
• statistical multiplexing
• decentralized architecture
• end-to-end principle

33
Next Lecture (contd.)

• Papers to read
• D. Clark, The Design Philosophy of the DARPA
  Internet Protocols, ACM SIGCOMM 1988
• .H. Saltzer, D.P. Reed, D.D. Clark, End-to-End
  Arguments in System Design, ACM TOCS, November
  1984
• The optional papers are also classics!
• No reviews due