EE499 Computer and Wireless Networking Laboratory

1
EE-499 Computer and Wireless Networking Laboratory

• Ahmed Helmy
• Spring 2002

2
Course Structure

• Three main components
• Lecture sessions/class participation
• Lab experiments
• Major Semester Project
• (check syllabus for more details)
• Web site ceng.usc.edu/helmy/ee499
• Student-centered, seminar-like, hands-on,
  thought-provoking, intro. Lab w/ advanced
  research course in networking!

3
Course Team
Assistants

• Jabed Faruque

• Sandeep Mudgal

Students - Around 10 groups, each of 3
students Sessions - 1 lecture session (3 hrs) -
2 lab sessions (3 hrs each) - Each group gets 1
lab session/wk
4
Milestones

• Forming groups for lab/project (1st 2 weeks)
• Paper reviews (almost weekly)
• Lab experiments (av. 1 new exp every 2 wks)
• Initial Project Proposal (5th week)
• Final Project Proposal (8th week)
• Initial Project Report (11th week)
• Final Project Report/demos (last wk of class)
• Class presentations (sign up)

5
Introduction and Motivation

• What Does the Internet Look Like?

6
100 node transit-stub topology
7
Map of the Mbone (3000 nodes)
8
Mbone map collected on Feb 23, 1999
Total routers 4,178 Total
interfaces 14,671 Total neighbors
17,798 (http//www.isi.edu/scan/mbone.html)
9
Small ISP (75nodes)
Medium ISP (500 nodes!)
Large ISP (3000 nodes!)
(http//www.isi.edu/scan/scan.htmlMercator)
10
Map of the Internet (50,000 nodes)
11
2 wks snapshot 2-15 Oct 00, 645k IP adds, 1M
links, 16 monitors probing 400k dest covering
48k routable prefixes (52), 7500 ASs
12
It is complex

• How do we measure the Internet?
• What do we measure?
• Topology, traffic, patterns of dynamics (e.g.,
  failures multicast membership), congestion
  (e.g., Q sizes), etc.

Mobility patterns, bandwidth as f(t)
How do we evaluate and test the existing
network? How do we simulate the Internet? How
do we understand and analyze the results? How
do we design protocols for the future? How do
we evaluate and test the future Internet??!! ...
13
Future Networks

• How does the future network look like?
• World turning wireless!
• Networked devices embedded more into our everyday
  life!
• Will this happen over night?
• Probably not
• There will be a long phase of coexistence with
  the wired world ...

14
Integrating Ad-Hoc Networks into the Internet
(Vision and Challenges)
The integrated architecture between the wired
Internet and ad hoc mobile networks enables
ubiquitous real-time apps and group
communication between users with different
mobility degrees, computing and communication
capabilities and different forms of network
connectivity. Architectural components include
efficient support for multicast and real-time
apps in wired and ad hoc networks, as well as
interoperability gateways.
15
Issues in the Integrated Design

• Objective to provide efficient real-time group
  communication in integrated ad-hoc Internet
  environment
• Main components
• Efficient mobility support for real-time apps
• Multicast support in ad-hoc networks
• Power-aware protocols for efficient ad-hoc nets
• Interoperability architecture

16
General Network Design Framework
17
Network Protocol Architecture Methodology

• Define the design space/domain parameters (the
  target environment)
• Design requirements
• Scale users, systems, sessions or calls
• Reliability (availability)
• Robustness (proper operation in presence of
  failure)
• Performance throughput, delay, jitter, overhead,
  etc.
• Environment
• Topology (LAN, WAN) and connectivity
• Characteristics of media
• wireless (high BER) vs fiber, mobile vs static,
  etc.
• Demand, traffic, applications

18

• Design determine initial parameters of the
  network or/and protocol
• Specification state/stipulate clearly, crisply
  and formally, the rules that govern the operation
  of the network or protocol
• Representation
• Finite state machine (FSM), pseudo code, English!
• Observation
• Much of the spec deals with failures/anomalies
• Most protocols (esp. network/mac layer) do not
  have clear robustness performance claims !
• How can we evaluate/test them?

19

• Evaluate the design
• Evaluation criteria
• Performance (e.g., overhead, response time,
  throughput)
• Correctness (e.g., absence of deadlocks or
  duplicates)
• Evaluation/modeling methodology
• Analysis (mathematical model) e.g. blocking/cell
  delay in 1 switch
• Simulation
• Hybrid e.g. of retransmissions of 100 TCP
  connections over 1000 node network

20

• Testing proof of concept mainly in testbeds
• Standardization (IETF RFCs, ATM forum, ITU/CCITT
• Deployment operation Interoperability
• Measurements, monitoring and management
• Modifications

21
Elements of Network Evaluation Studies

• Evaluation metrics
• correctness, performance need clear definition
• Evaluation Methodology
• Analytical (queuing theory)
• Network simulation (e.g., VINT/NS)
• FSM search (e.g., STRESS)
• Experimentation/measurements
• Analysis of results and conclusions

These are extremely important elements to
define for the projects
22
Potential Research Directions in Computer Networks
Protocol Design and Testing Tools
Mobile and Ad Hoc Networks
Network Simulation (VINT/NS) USC/ISI
Scalable Multicast in Ad Hoc Networks
Automatic Protocol Test Synthesis (STRESS)
Efficient Mobility/Handoff support using
Multicast-based Mobility (MM)
23
Efficient Mobility/Handoff Support Using
Multicast-based Mobility (MM)
CH
CH Correspondent host (sender)
Wireless link
Mobile Node
Distribution tree dynamics while roaming
24
MM Issues

• Ubiquitous deployment
• need inter-domain multicast
• Scalability
• multicast state kept in routers will grow
• Management
• global multicast address allocation
• Promising direction to pursue
• intra-domain multicast with state aggregation

25
Multicast in Large-scale ad hoc networks

• Multicast service has the rendezvous problem
• broadcast/prune does not scale
• sparse mode need bootstrap mech. to find RP
• hierarchical multicast needs domains/ASs
• Need simple, scalable hierarchy formation that
  adapts to dynamics
• Need to bootstrap multicast service w/o ASs
• Geographic multicast address allocation
• Adaptive query/response mechanisms for routing

26
Overview of Simple Hierarchy Formation
ZRP zone-based routing protocol
27
(No Transcript)
28
(No Transcript)
29
Issues with contact ZRP

• How do we choose the contacts?
• Use mobility profiles, capabilities, etc.
• What are the characteristics of the resulting
  graphs? (this is function of time) Are they
  really small world? How long do they stay this
  way?
• This is subject to mobility and traffic patterns
• What is the advantage/overhead of this scheme
  (vs. others)?

30
Query-response Mechanism with Geographic Address
Allocation
31
Issues with Geo Address Allocation

• What is the mapping function?
• Can we model this problem as a peer-to-peer
  network (e.g., CAN or Chord)? This will be
  dynamic, however.
• What if the RR does not have nodes?
• Need a fallback mechanism

32
Adapting to Dynamics in Membership
33
Issues with adaptive anycast

• How do we promote local servers?
• What is the gain/overhead?

34
The Virtual InterNetwork Testbed (VINT)
- Developed at USC/ISI - Used to evaluate and
simulate varieties of network protocols mixes -
Provides composable simulation environment -
Supports animation and visualization of networks
and protocols - Provides simulation abstraction
techniques for scalability - Includes an
extensive library of routing, transport,
multicast and other protocols - Recently
augmented with support for mobility and power
modeling
35
Network Simulation Issues

• Scaling simulations to (millions) of nodes!
• Abstracting details
• Parallel simulations
• Mixed mode simulations
• Scenario generation
• Q. Which scenarios do we simulate and why?
• Q. What can we say about quality of simulation
  when were done?
• Need automatic/systematic scenario generation

36
STRESS Systematic Testing of Robustness by
Evaluation of Synthesized Scenarios
- STRESS provides algorithms and tools for
systematic and automatic synthesis of stress
scenarios - stress scenarios expose protocol
design errors or drive the protocol into
worst-case performance - Synthesized scenarios
include topology description, sequences of events
and faults (or network dynamics) - These
scenarios may be later simulated or applied to
actual implementation of a protocol
37
Stress Issues

• How to model the protocol/system under study?
• Use FSM and GFSM with extended semantics
• How to synthesize scenarios? Using search?
• What about search complexity?
• Do we have to use forward search?
• Goal oriented backward search seems very
  promising
• Applicability mcast rtg, ad hoc mac, rel. mcast,
  mcast cong ctl, ad hoc rtg, MIP, ...

38
Future Work

• Lots of interesting research issues
• Identify and relate to some of them
• Start formulating some ideas for the project
• Lab experiments will increase your skills and
  help develop insights (hopefully!)
• More to come...

39
Lab (briefly)

• Develop basic skills to setup a network testbed
  (setup static/dynamic routing)
• Use tools to measure and observe the network
  parameters e.g., topology, traffic, delays
  (ping, traceroute, tcpdump, etc.)
• Observe characteristics of wireless networks
  (physical, MAC, MIP, TCP/wireless)
• Analyze collected data in a meaningful way to get
  sound conclusions
• Chk web for more details

40
References

• Class web page ceng.usc.edu/helmy/ee499
• My web page ceng.usc.edu/helmy
• Others
• catarina.usc.edu/vint
• catarina.usc.edu/stress
• catarina.usc.edu/pim