Performance Analysis of a CSMACD Bus Network

1
Performance Analysis of a CSMA/CD Bus Network
CS4554 Network Modeling and Simulation
by

• Team Foxtrot!
• Maj Jim McDonald
• Capt Rod Lewis
• Capt Mike Del Grosso

18 Sept 2000
2
Presentation agenda

• Define Problem
• State Goal
• Assumptions
• Network Description
• Performance Metrics
• Workload
• Analytical Model of CSMA/CD Network
• Simulation Demonstration
• Conclusions
• Lessons learned

3
Define Problem
A very common network today is the CSMA/CD or
Ethernet network. It is important to establish
the performance characteristics of networks
before putting them to use. This will assist in
keeping them from becoming overloaded so they
provide the best service possible to its users.
Analyzing and modeling these networks can also
help plan for network growth and adjustments.
4
Goal Statement

• Complete an analytical model and a simulation
  model of a CSMA/CD (Ethernet) network and compare
  the evaluation techniques results.

5
Assumptions

• When one or more bits overlap in time all are
  lost and must be retransmitted.
• All nodes listen constantly to the bus and only
  transmit if there is no transmission already on
  the bus.
• If there is no transmission on the bus any node
  with available data can transmit immediately
• System performs the service correctly

6
Bus Network

• Description of Simulated Bus Network

7
Performance Metrics

• We used a simplified example of a Bus Network for
  our project, which includes
• 1. Performance is measured by the time taken to
  perform the service
• 2. The rate at which the service is performed
• 3. The resources consumed while performing the
  service
• SMART - Specific, Measurable, Acceptable,
  Realizable and Thorough

8
Performance Metrics

• Speed
• The access delay at any station should be less
  than 1 second
• Sustained throughput must be at least 10
  Mbits/sec
• Reliability
• The probability of any bit being in error must be
  less than
• The probability of any frame being in error must
  be less than 1
• The probability of any frame in error being
  delivered without error indication must be less
  than

9
Performance Metrics

• Availability
• The mean time to initialize the LAN must be less
  than 15 milliseconds.
• The mean time between LAN initializations must be
  at least 1 minute.
• The use of the analytical model checks the
  realizability of all our numerical values.

10
Test Workload

• System Under Test
• 1.There are 10 nodes attached to the network.
• 2. Bus length is 40 meters.
• 3. Fixed packet length. 2.
• Workload Areas of OSI 7 layer Model
• Datalink Layer - transmission of frames over a
  single link
• Network Layer - distance between source and
  destination and the characteristics of packets
  transmitted
• Transport Layer - specifying the frequency sizes
  and other characteristics of various messages

11
Test Workload

• Loading Level - is of a typical case
• (not full capacity or worst case)
• No significant impact of external components.
• This workload can be easily repeated.

12
Analytical Model

• Type of network analyzed - M/M/1
• Inter-arrival times and Service Times are
  exponentially distributed and there is one server
• State transition diagram for an M/M/1 queue

...
13
Simulation ModelPlanning

• SMART Goal
• Level of detail is identified
• Background of team
• some programming, management level knowledge,
  eager
• Time - big issue

14
Simulation Model

• Random number generator checked
• Uniformity and Independence
• Results reviewed on a constant basis
• Documentation given in the form of final report
• Length of time run was appropriate
• V and V...

15
Validation and Verification

• Coming from the the phrase until validated,
  all evaluation results are suspect
• Use the rules
• Do not trust the results of a simulation model
  until they have been validated by an analytical
  modeling or measurements.
• Do not trust the results of an analytical model
  until they have been validated by a simulation
  model or measurements.

16
Analytical Model

• Mean service time
• Mean arrival rate
• Total arrival rate

17
Analytical Model

• Traffic intensity
• End-to-end propagation delay
• For constant packet lengths

18
Analytical Model

• Mean delay

19
Analytical Model
20
Analytical Model
21
Analytical Model
22
Analytical Model
23

Simulation Model

• Use the OPNET modeler to simulate an M/M/1/B
  queue
• Validate the results with an analytical model
  (M/M/1/B)

The goal of the study is to evaluate and compare
the CSMA/CD Bus Network with an acceptable loss
rate.
24
Profile Setup
25
Bus Utilization
26
WorkStation Delay
27
Load (packet/sec)
28
Delay(sec)
29
Bus Load
30
Network Delay
31
Validation and Verification

• Delay differences
• Inaccurate propagation delay b/n analytical and
  simulation
• Definition of Light and Heavy unknown.
• Simulation is unknown
• Analytical model is hard numbers

32
Conclusions

• The initial problem was strongly bound so that
  analytical modeling of the proposed network could
  be performed with available tools.
• Analytical modeling produced acceptable solutions
  for the given problem and considering stated
  assumptions.
• Opnet Modeler is a very powerful simulation tool.
  It also requires extensive training to be able to
  produce correct and useful models.

33
Lessons Learned

• Follow the directions in the readings!
• You cant jump into a simulation tool or analyze
  someones code within a very constrained time
  frame.
• Need more time to learn simulation tool given.
• Focus and scaling of the problem/exercise.

34
Questions?
35
Thank You