Upload
earl-anderson
View
226
Download
5
Tags:
Embed Size (px)
Citation preview
1
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
ModelerModelerDay 3Day 3
2
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Course ContentDAY 3:
• Configuring Standard Applications {LABs}
• Traffic Modeling Techniques {LABs}
• Modeling Large Networks
• Importing Topology and Traffic
DAY 4:
• Process Modeling Methodology
• Accelerating Wireless Simulations{LABs}
• Custom Wireless Effect {LABs}
• Debugging Simulation Models{LABs}
3
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Configuring Configuring Standard Standard
ApplicationsApplications
4
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Agenda• Application basics
• Network traffic (users, devices, networks, etc.)• Modeling traffic in OPNET• Application modeling techniques• OPNET application architecture
• Configuring applications workflow• Configure applications• Define profiles• Configure server/peer• Deploy profiles• Lab 1: OPNK2003 cyber café
• Configuring analytical traffic on client-client applications• Concepts and attributes• Lab 2: VoIP configuration
• Easy configuration of application traffic• Concepts and attributes• Lab 3: application demands
• Troubleshooting• Lab 4: troubleshooting (bonus)
5
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Take-Away Points
• Standard applications available
• Application configuration
• Defining user profile
• Client & server setup for standard applications
• Application demands
• Application modeling techniques
• Application troubleshooting
6
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Agenda• Application basics
• Network traffic (users, devices, networks, etc.)• Modeling traffic in OPNET• Application modeling techniques• OPNET application architecture
• Configuring applications workflow• Configure applications• Define profiles• Configure server/peer• Deploy profiles• Lab 1: OPNK2003 cyber café
• Configuring analytical traffic on client-client applications• Concepts and attributes• Lab 2: VoIP configuration
• Easy configuration of application traffic• Concepts and attributes• Lab 3: application demands
• Troubleshooting• Lab 4: troubleshooting (bonus)
7
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
How Do We Define Network Traffic?
• Source (user, group of users)
• Destination (user, server)
• Traffic patterns
8
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Applications
• Client-server applications
• Examples: HTTP, FTP
• Peer-peer applications
• Examples: voice, video
• Multi-tier applications
• Examples: web-based applications, e-commerce, home-grown applications
REQUEST
RESPONSE
RESPONSE
3
RESPONSE
4
REQUEST
1
REQUEST
2
REQUEST
RESPONSE
REQUESTRESPON
SE
9
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Application Modeling Techniques
• Explicit• Simulate all packets in detail
• Very high fidelity results
• Potentially longer run times
• Analytical• Simulate majority of traffic using mathematical representation
• Faster simulation run times
• Hybrid• Combination of both explicit and analytical modeling techniques
10
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
OPNET Application Architecture
PhasesD e fine com m u n ica tio n p a tte rns w ith in ea ch ta sk
TasksD e sc rib e b as ic un it o f u se r ac tiv ity
Applica tionsD e fine a pp lica tio n co n fig u ra tion
Pro filesD e scrib e use r b e h av io ur
Custom application
11
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Profile and Applications: Example
H T T P F T P
E n g in e er
V id eo H T T P
M a rke tin g D e p a rtm e nt
V o ice D a tab a se
S a le s E n g in e erProfiles
Apps
12
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Configuring Application Models: Workflow
Configure applications
Define profiles
Configure servers/peers
Deploy profiles
13
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Agenda• Application basics
• Network traffic (users, devices, networks, etc.)• Modeling traffic in OPNET• Application modeling techniques• OPNET application architecture
• Configuring applications workflow• Configure applications• Define profiles• Configure server/peer• Deploy profiles• Lab 1: OPNK2003 cyber café
• Configuring analytical traffic on client-client applications• Concepts and attributes• Lab 2: VoIP configuration
• Easy configuration of application traffic• Concepts and attributes• Lab 3: AppDemands
• Troubleshooting• Lab 4: troubleshooting (bonus)
14
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Configure applications
Configuring Applications
Define profiles
Configure servers/peers
Deploy profiles
15
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Configure Applications
16
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Applications: Built-in Common Applications
Custom Application
Name Description
Database
FTP
HTTP
Remote Login
Voice
Video Conferencing
Generic/customizable application
Data entry/query application
E-mail application
File transfer protocol application
Web-browsing application
Print job
Telnet application application
Client-to-client voice application
Client-to-client video application
Standard applications
17
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Applications
• Session of specific request/response patterns
• Commonly used applications: email, http, ftp…
• Common settings to all applications• Type of service
• Symbolic server name
• RSVP parameters
• Specific settings to each application• HTTP page configurations
• FTP file sizes
• Database query ratio
18
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Application Definition Utility
• Contains all application definitions• Reusability
• Across clients• Across scenarios
• Easier deployment• Easier change
II’m running “FTP (Light)”
II’m running “FTP (Light)”
II’m running “FTP (Light)” I
I’m running “FTP (Light)”
II’m running
“FTP (Light)”
II’m running
“FTP (Light)”
“FTP (Light)” Application Definition
19
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Symbolic Server Name
Application has a symbolic server name, e.g. “FTP (Light)”
Actual server name is specified in server, e.g. WashingtonDC_Server
Mapping between symbolic server name and actual server name
20
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Application Definition Object
21
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Example: File Transfer Protocol• Upload or download of files:
• Using file size distribution
• Using Inter-request time distribution
Common settings
Specific FTP settings
FTP (Application)
PUT PUT GET PUT
Inter-Request
Time
Inter-Request
Time
Inter-Request
Time
time
22
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Example: Hypertext Transfer Protocol (HTTP)
• Download of a web page:• Using distribution for number of objects in page and sizes
• Using page interarrival time distribution
Common settings
Specific HTTP settings
HTTP (Application)
Web page download
Page Inter arrival Time
Page Inter arrival Time
timeWeb page download
Web page download
23
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Hypertext Transfer Protocol (cont.)
24
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Define profiles
Profiles
Configure applications
Configure servers/peers
Deploy profiles
25
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Define Profiles
26
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Profiles
• Profile = Reusable collection of applications• Describes activity patterns of:
• an individual user• a group of users
Engineer
CEOCFO
Engineer
Engineer
Help desk
Marketing department
27
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Profile Configuration Object
• Contains all profile definitions• Reusability
• Across clients• Across scenarios
• Easier deployment• Easier change
II’m an
engineer
II’m an engineer
II’m an engineer
II’m an engineer
II’m an engineer
II’m an engineer
Engineer Profile
Definition
28
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Profile Properties
Profile Name
Applications
Operation Mode
Start Time
Duration
Repeatability
29
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Profile Name, Start Time, and Duration
Engineer
9:00am 12:00pm 5:00pm
Start Time
Duration
• Example:• “Engineer” profile• Starting at 9:00am• Lasting 8 hours
30
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Profile Repeatability
• Serial mode
Engineer Engineer Engineer EngineerInter-repetition time Inter-repetition time Inter-repetition time
• Concurrent mode
Engineer
Engineer
Engineer
EngineerInter-repetition time
Inter-repetition time
Inter-repetition time
31
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Applications in Profile
I have the definitions of the following applications:• Database Access (Light)• Email (Light)• Web Browsing (Light)• FTP
32
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Application Scheduling
“Engineer” profile
Web Browsing
9:00am 17:00pm
Web Browsing Web Browsing
FTP FTP FTP FTP FTP
Database
12:00pm
• Multiple applications per profile
33
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Operation Mode
• Simultaneous
Web Browsing
FTP
Database
9:00am 12:00pm 17:00pm
• Serial (ordered/random)
EmailWeb BrowsingDatabase
9:00am 12:00pm 17:00pm
FTP
34
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Applications in Profile
Name
Start Time Offset
Duration
Repeatability
35
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Name, Start Time Offset, and Duration
• Example:• “FTP” application• Starting 2 hours after profile “Engineer” starts• Lasting 4 hours
DurationStart time offset
9:00am 5:00pm11:00am 3:00pm
Note: an application will not last longer than the profile
PUT PUT GET PUT
FTP (Application)
Engineer (Profile)
36
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Application Repeatability
• Serial mode
FTP FTP FTP FTPInter-repetition time Inter-repetition time Inter-repetition time
• Concurrent mode
FTP
FTP
FTP
FTPInter-repetition time
Inter-repetition time
Inter-repetition time
Note: an application will not repeat and last longer than the profile
37
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Configure servers/peers
Configure Servers/Peers
Configure applications
Define profiles
Deploy profiles
38
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Configure Server/Peers (cont.)
39
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Server Supported Service Setup
40
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Server CPU Setup
41
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Server Address
• Uniquely identifies server
• Server address to be mapped to the symbolic server name
42
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Deploy profiles
Deploy Profiles
Configure applications
Define profiles
Configure servers/peers
43
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Deploy Profiles (cont.)
44
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Client Profile Setup
45
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Destination Preferences
• No destination preference => Randomly select among destinations supporting service• Selection weight set on destinations
Explicitly configure destination(s) => Randomly select among destinations specified on client Weight set on client
Server1
Server2
Client
Server2 is 5X more likely to be chosen
?
Server1
Server2Client
Server1 is 2X more likely to be chosen
?
I have to choose between: - Server1: weight 20 - Server2: weight 10
My weight is 10Choose me!
My weight is 50Choose me!
46
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Configuring Destination Preferences
Application refers to a symbolic server name, e.g. web_browsing_server
Actual server name is specified in server, e.g. WashingtonDC_Server
Mapping between symbolic server name and actual server name
47
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Configuring Destination Preferences (cont.)
48
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Application Statistics
• Local and global statistics• Client/Server statistics or
calling/called party statistics
• Specific to each application
• Local statistic results will be grouped per profile/application
49
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Lab 1: OPNK2003 Cyber Cafe• Objective
• Apply workflow to configure application traffic
• Application configuration
• User behavior profiles
• Configure servers to support applications
• Deploy user profiles
• Refer to lab handout “Lab1: HTTP Application Configuration Lab” on page 1
Configure applications
Define profiles
Configure servers/peers
Deploy profiles
50
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Agenda• Application basics
• Network traffic (users, devices, networks, etc.)• Modeling traffic in OPNET• Application modeling techniques• OPNET application architecture
• Configuring applications workflow• Configure applications• Define profiles• Configure server/peer• Deploy profiles• Lab 1: OPNK2003 cyber café
• Configuring analytical traffic on client-client applications• Concepts and attributes• Lab 2: VoIP configuration
• Easy configuration of application traffic• Concepts and attributes• Lab 3: AppDemands
• Troubleshooting• Lab 4: troubleshooting
51
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Configuring Analytical Traffic on Voice and Video Applications• Voice and video: intensive discrete-event applications
• Streaming applications
• Each small packet is modeled
• Long simulation run
• Not scalable
• Solution to improve simulation performances• Analytical traffic
• Interpret all traffic as a background traffic flow
• Loss of accuracy
• Hybrid traffic
• Model part of the discrete traffic in an analytical flow
• Keep accurate response time and jitter
• Still model specific internals of voice application
52
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Hybrid Traffic ConceptDiscrete event traffic
Analytical traffic
Hybrid traffic
53
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
VoIP “Traffic Mix” Attribute
54
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Video Conferencing “Traffic Mix” Attribute
55
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Methodology
• Configure most of your VoIP or video nodes with background traffic
• Configure the node of interest to use discrete traffic
56
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Lab2: Configuring VoIP
• Objective• Apply hybrid traffic to generate VoIP traffic and get end-to-end delay
• Refer to lab handout “Lab2: VoIP Application Configuration Lab” on page 26
57
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Agenda• Application basics
• Network traffic (users, devices, networks, etc.)• Modeling traffic in OPNET• Application modeling techniques• OPNET application architecture
• Configuring applications workflow• Configure applications• Define profiles• Configure server/peer• Deploy profiles• Lab 1: OPNK2003 cyber café
• Configuring analytical traffic on client-client applications• Concepts and attributes• Lab 2: VoIP configuration
• Easy configuration of application traffic• Concepts and attributes• Lab 3: AppDemands
• Troubleshooting• Lab 4: troubleshooting (bonus)
58
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Application Demand Definition• Abstract of a client-server or client-client application
• Connects two end-nodes• Request/response oriented configuration• Available in “applications” object palette
• Advantages• Configuration friendly• Custom-made traffic• Possibility to use discrete, hybrid, or analytical traffic• Visual representation of the application flows
• Limitations• No specific internals of a standard application• Only for simple traffic patterns
59
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Application Demand Concepts
Client or
server
Client or
server
Requests
TimeStart Time
End Time
Rate (requests/hour)
Responses
60
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Application Demand Deployment
• Click and connect• No need for configuration objects
• Right-click on demand to configure
• Demands can be copied and pasted for duplication
61
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Application Demand Attributes
• Generic traffic configuration• Duration• Request parameters• Response parameters
• Hybrid traffic• Transport protocol
62
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Application Demand Deployment Wizard• Faster deployment to create application demands
• Full mesh
• From one node to many nodes
• From many nodes to one node
• Note:• If 2 or more nodes are selected before starting wizard, configuration applies to selected nodes only• If no node is selected before starting wizard, configuration applies to all nodes in network
63
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Application Demand Statistics
• Statistics per demand collected on end-nodes• Traffic sent/received
• Response time
• No global statistics
64
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Lab3: Deploying Application Demands• Objective
• Generate traffic quickly using Application Demands
• Please refer to lab handout “Application Demand Configuration Lab”
65
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Agenda• Application basics
• Network traffic (users, devices, networks, etc.)• Modeling traffic in OPNET• Application modeling techniques• OPNET application architecture
• Configuring applications workflow• Configure applications• Define profiles• Configure server/peer• Deploy profiles• Lab 1: OPNK2003 cyber café
• Configuring analytical traffic on client-client applications• Concepts and attributes• Lab 2: VoIP configuration
• Easy configuration of application traffic• Concepts and attributes• Lab 3: AppDemands
• Troubleshooting• Lab 4: troubleshooting (bonus)
66
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Troubleshooting Guidelines• «I don't see any traffic» or «I don't see expected traffic»
• Check simulation logs
• Look at the “Common problems” section of this presentation
• Use constant distributions to make profile scheduling deterministic
• Collect application Traffic Sent/Received stats in “All Values” mode
• Shows exact size and time data was sent
• Make sure this is not a network issue
• Connect an application demand between your client and server
• See if you get application demand response time statistics
• Scale down the network
• Isolate the problem
• Use “User Defined Reports” in “Scenarios” menu
• Globally visualize attribute settings
• Use “Network Difference Report” in “Scenarios” menu
• Compare to a similar working scenario
67
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Troubleshooting: Simulation Logs
• Simulation = first place to look at
• Available in “DES” menu
68
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Troubleshooting: Simulation Logs (cont.)
“Engineer” profile
“FTP” application “HTTP” application
Simulation
“Engineer” profile
Simulation
“Engineer” profile
Profile won't repeat if profile duration is equal to simulation duration.
In serial mode, an application will not start if previous application duration is equal to profile duration.
69
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Troubleshooting: Simulation Logs (cont.)
• Profile set on a client doesn't exist
Service set on server doesn't exist
“Engineer” profile
?
?“FTP (Light)” application
I’m an engineer
Sorry, I don’t know what an engineer is.
I’m running “FTP
(Light)”
Sorry, I don’t know what “FTP (Light)” is.
70
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Troubleshooting: Simulation Logs (cont.)
• All required application parameters are not configured
“FTP (Light)” is a FTP application
Sorry, I need more information to use “FTP (Light)”:
• File size
• Inter-request time
Service set on server doesn't exist
?Profile
“engineer” uses “FTP (Light)”
Sorry, I don’t know what “FTP (Light)” is.
71
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Common Problems ‘Profile start time + application start time offset’ < ‘Convergence time of routing protocols
and spanning tree algorithm etc’
I want to send this FTP packet
Wait, I don’t have a route to the server yet
‘Profile start time + application start time offset’ > ‘Simulation end time’ Common when using exponential distribution with start time and/or start time offset
“Engineer” profile
Simulation
“FTP” application
“Engineer” profileSimulation
Profile start time
Profile start time Application start time offset
“FTP” application
72
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Common Problems (cont.) If application duration is longer than profile duration, application will
be terminated when profile is over.
The symbol map “Unlimited” for profile and application “Repeatability” has a default interarrival time of 300 secs by default.
“Engineer” profile
“FTP” application
Simulation
“Engineer” profile
“Engineer” profile
“Engineer” profile
“Engineer” profile
300 secs 300 secs 300 secs
Etc…
73
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Common Problems (cont.)
Voice and video conferencing can be configured between two clients/LANs only.
Applications other than voice and video conferencing can be configured between a client/LAN and a server/LAN only.
Voice/Video
FTP, email, database, http,
remote login, print
wrong
wrong
74
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Advanced Troubleshooting
• Diagnostic traces• Per application type: “ltrace gna_<application_type_name>”
Examples: “ltrace gna_email”, “ltrace gna_ftp”, “ltrace gna_http”• For all applications: “ltrace gna”• Simple example
• Start simulation under ODB • ODB> ltrace gna• ODB> cont
• This is an advanced technique. For more details, see session 1502: Debugging Simulation Models — Introduction
• Run OPNET debugger (ODB)
75
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Lab4: Troubleshooting
• Objective• Learn how to troubleshoot an application setup
• Please refer to lab handout “Application Troubleshooting Lab”
76
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Configuring Custom Application: Workflow
Define Profiles
Configure tiers
Deploy profiles
Configure tasks
Configure custom app
Break down the application
77
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Breakdown of the Application• Identify the tiers
For example, an online e-commerce application might involve• Client• Web server• Authentication server• Database server
78
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Summary
• Standard applications available
• Profile configuration
• Application configuration
• Client & server setup for standard applications
• Application demands
• Application modeling techniques
• Troubleshooting
79
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Traffic Modeling Traffic Modeling TechniquesTechniques
80
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Outline
• Motivation for traffic modeling
• Traffic representation
• Traffic generation: sources and import methods
• Traffic modeling: comparison of various OPNET approaches
• Lab #1• Compare accuracy and speed of explicit traffic simulation
with background traffic
• Lab #2• End-to-end delays using hybrid simulation
81
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Motivation
• It is critical to provide good inputs (garbage in → garbage out)• network topology
• traffic
• It is equally critical to choose a traffic modeling technique• simulation speed vs. accuracy
• This session focuses on modeling traffic
82
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Outline
• Motivation for traffic modeling
• Traffic representation
• Traffic generation: sources and import methods
• Traffic modeling: comparison of various OPNET approaches
• Lab #1• Compare accuracy and speed of explicit traffic simulation
with background traffic
• Lab #2• End-to-end delays using hybrid simulation
• Lab #3• Flow Analysis
83
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Real-world Traffic Representation• Types of traffic
• WAN, LAN, application traffic
• Traffic representation in OPNET
• Choice of representation depends on modeling purpose• Packet by packet
• End-to-end delays, protocol details, segmentation effects
• Aggregated traffic
• Capacity planning, steady-state routing analysis
Traffic Flows, Device/Link Loads
(Background Traffic)
Aggregated Traffic
Explicit TrafficPacket by Packet
OPNET RepresentationTraffic Type
84
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Parameters in Traffic Representation
• Traffic Parameters• Packet inter-arrival time or arrival rate
• seconds, packets/second
• Packet size
• bits, bytes etc.
• bits/sec = bits/pkt * pkts/sec
• Variability settings• Probability Distribution Function (PDF)
• standard (exponential, bernoulli etc)
• custom
• build your own with PDF editor
• scripted
Interarrival time
File Size
85
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Outline
• Motivation for traffic modeling
• Traffic representation
• Traffic generation: sources and import methods
• Traffic modeling: comparison of various OPNET approaches
• Lab #1• Compare accuracy and speed of explicit traffic simulation
with background traffic
• Lab #2• End-to-end delays using hybrid simulation
86
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Sources of Explicit Traffic• Explicit traffic injected at
• Application Layer
• email, HTTP, FTP etc
• ACE, app_demands
• Network Layer
• RPG (self-similar traffic generator)
• Lower layers
• Native protocol sources
• Ethernet, ATM, Frame Relay etc
87
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Sources of Aggregated Traffic• Traffic flows injected at
• Application Layer
• app_demands
• Network Layer
• IP traffic flows
• Lower Layers
• ATM traffic flows
• Element loads• CPU utilization
• Link loads
• Do not require source models
IP Traffic Flow
app_demand
88
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Traffic Data Import
Traffic Data
Traffic Data
Network monitoring software samples traffic periodically using probes
Export to text files or OPNET recognizable formats
Import into OPNET
89
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Importing Explicit Traffic – Packet Trace
• Packet traces captured using network analyzer• Sniffer analyzer
• tcpdump
• windump
• Application Characterization Editor (ACE)
90
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Importing Server Data• Import from
• HP OpenView Performance Agent
• BMC Patrol• Concord SysEDGE• NetIQ AppManager• Generic XML
• Server Characterization Editor (SCE)
91
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Importing Aggregated Traffic – Flow Data• Flow-specific data • Demand objects
92
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Importing Aggregated Traffic – Link Loads
• Link load information from• Concord NetworkHealth
• MRTG
• Spreadsheet (text info)
• Can be converted into traffic flows
93
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Outline
• Motivation for traffic modeling
• Traffic representation
• Traffic generation: sources and import methods
• Traffic modeling: comparison of various OPNET approaches
• Lab #1• Compare accuracy and speed of explicit traffic simulation
with background traffic
• Lab #2• End-to-end delays using hybrid simulation
94
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Simulation Techniques
• Discrete event simulation• Model all traffic (data, signaling, management) using packets
• Account for all timers in every protocol layer
• Perform every state/event transitions of all protocol layers
• Analytical simulation• Abstract queue performance using mathematical equations
• Model traffic as state information in various network elements
• Hybrid simulation• Mix of modeling approaches (discrete event + analytical)
• Mixture of traffic types (explicit traffic + aggregated traffic)
95
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Comparison of Various Simulation Techniques
96
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Examples of Analytical Methods
Used by mac layer in LAN models
Used by IP CPU
97
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Analytical Simulation
• Delay estimation
• Capacity planning
• Routing analysis
• QuickPredict
• Flow Analysis performs analytical simulations for steady-state analysis
98
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Comparison of Various Simulation Techniques
99
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Example of Discrete Traffic Simulation
• TCP congestion window management
• End-to-end delays at application layer
Application
Presentation
Transport
Network
Data Link Layer
Physical
Session
Application
Presentation
Transport
Network
Data Link Layer
Physical
Session
Congestion Window
E2E Delay
100
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Discrete Event Simulation
• Packet retransmission due to varying acknowledgment delays
• TCP/ATM connection setup delay for varying congestion levels of a network
101
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Comparison of Various Simulation Techniques
102
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Micro-Simulation – Delay Estimation
• Delay estimator (t) for explicit packets due to additional aggregated traffic contending for a common resource (such as a link)
• Retrospective micro-sim as lazy evaluation on real packet arrival
• Sensitive to queuing schemes like PQ, CQ and WFQ
r1 r2
t
b1 b2 b3
Time
Micro-simulation
103
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Analytical Coupled with Micro-Simulation
r1 r2
t + a
b1 b2 b3
Time
Micro-simulationAnalytical-simulation
a
When real packets are separated by a long intervalLazy evaluation in two stages: analytical + micro-simulation
Currently implemented for FIFO queues onlyGain in simulation speed compared to pure micro-simulation
104
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Hybrid Simulation
105
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Outline
• Motivation for traffic modeling
• Traffic representation
• Traffic generation: sources and import methods
• Traffic modeling: comparison of various OPNET approaches
• Lab #1• Compare accuracy and speed of explicit traffic simulation
with background traffic
• Lab #2• End-to-end delays using hybrid simulation
106
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
• Create a simple network with explicit traffic and run a simulation
2. Duplicate the network and replace the explicit traffic with corresponding background traffic
• Run the simulation with background traffic
• Assess/compare the accuracy and speed of the two traffic modeling approaches
• Refer to lab manual for instructions
Lab 1: Overview
107
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Destination nodes
Source nodes (video servers) with different ToS (1,2,3) Interface with WFQ
100 video users (MPEG4) - 3 available service classes
Queuing delays ?
Lab 1: Network Model
108
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Aggregated Traffic – Just a Number to Compute a Delay• Aggregated traffic in OPNET
• Traffic flows (routed background traffic)
• Device/link loads (static background traffic)
• Common to both type of aggregated traffic• Internal representation: just a number for a given node/link
• Effects: only one effect – to produce a delay
• One difference between the two types of aggregated traffic• Traffic flows (routed background traffic) needs to be propagated to each
node in the flow path tracer packets (propagators)
2Mbps
2Mbps
2Mbps
2Mbps
2Mbps
2Mbps
2Mbps
2Mbps
2Mbps
2Mbps
2Mbps
2Mbps
109
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Defining Traffic Flows
• 3 basic steps
110
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Defining Traffic Flows (cont.)
• Each flow is broken down into intervals as specified
Interval : 300-600 seconds
111
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
I1 I2 I3 I4 I5
Simulating Traffic Flows
• Refresher tracer packets
I1
P1 P3 P4 P5P2
112
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Simulating Traffic Flows (cont.)
• Sending new information on traffic level changes
start time = 150.00.0 I1 I2 I3 I4 I5
113
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Effects on Traffic Flows
• Clipping• Effect: to reduce traffic flows to within bottleneck capacities
• Modeled as: state information on each node, bps and pps rates inside tracer packet diminished downstream of the bottleneck
• Segmentation:• Effect: to reduce packet size to within MTU
• Modeled as: bps and pps rates inside the tracer adjusted to account for overhead due to segmentation
• Load Balancing:• Split traffic flows proportionally among available paths
• Modeled as: create copies of tracer packets with modified bps and pps rates
114
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Load Balancing Traffic Flows
0.75
0.25
0.5 * 0.75 = 0.375
0.375
0.25
0.25
0.3750.375
115
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Outline
• Motivation for traffic modeling
• Traffic representation
• Traffic generation: sources and import methods
• Traffic modeling: comparison of various OPNET approaches
• Lab #1• Compare accuracy and speed of explicit traffic simulation
with background traffic
• Lab #2• End-to-end delays using hybrid simulation
116
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Lab 2: Hybrid Simulation - Overview
• Video Conferencing end-to-end delays
• WAN – with background traffic
• Create one explicit flow – hybrid simulation
• Observe dependence of the explicit traffic end-to-end delay on Type of Service (ToS) used
117
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Lab 2: Network Model
• WAN with many background traffic flows• One explicit video traffic flow from Seattle Houston
(130 kbps – 360 kbps)• End-to-End delays ?• Refer to lab manual for instructions
118
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
ModelingModelingLarge NetworksLarge Networks
119
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Agenda• Conceptual goals
• Importing network topologies
• Choosing aggregation levels
• Using LAN models
• Importing traffic
• Using routed background traffic
• Managing projects and scenarios
• Navigating Large Topologies
• Controlling Simulation Runtime
• Effectively Using Background Traffic
• Software goals• Import topology
• Question and answer browser
• Import traffic
• Network browser
• Traffic browser
• Efficiency Modes
• Static Background Traffic
• Event Speed Parameter
• Specifying Traffic Growth
• Scheduling Automated Simulations
• Viewing and Editing Scheduled Simulations Log
120
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Overview – Large networks
• The concept of a “large” network does not necessarily refer to the number of nodes or the geographic extent of a network.
• A large network is any model that will generate a significant amount of events. This translates into longer simulations and memory limitation issues.
• We will explore methods for increasing the efficiency of your models when the economies of scale become an issue.
121
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Common Pitfalls
• OPNET is powerful software that tempts users with “grand designs”.
• Common Pitfall: “I will model everything”
• Developing models without clear direction and a solid understanding of the questions that need answering leads to inefficiency, and results that aren’t useful.
122
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Questions Should Drive Model Design
• It essential to first determine which questions to answer.
• Design the model to answer those questions. Don’t try to answer everything.
• A key factor in effective modeling is flexibility. Realize that over the course of the modeling process that the questions may change.
123
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
LAN Models
• Single nodes may be used to model entire LANs.
124
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Why Use LAN Models?• LAN models are a powerful
mechanism for abstracting complex local area networks that may contain multiple nodes with the same configurations.
• LAN models reduce clutter in the workspace.
• LAN models use less memory.
• LAN models generate fewer simulation events allowing simulations to run faster.
These models represent the same network.
125
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
• LAN Models behave as if they were specific nodes modeled using OPNET.
• The key difference is that you can collect statistics on the LAN models but not on the specific nodes and links within the model.
Additional Information About LAN Models
126
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Modeling Techniques
• Design Accurate, Efficient Models
• Two Techniques• Approximation
• Reduction
127
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Approximation
• Tradeoff: Precision vs. Performance
• Loss in precision means higher variance
• Overall results are still accurate
• Approximation Approach
• Apply a combination of analytically and explicitly modeled traffic
• Allows you to limit the number of variables, while retaining the proper traffic behavior for the model
128
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Traffic Types
• Three different types of traffic • Loads
• Also called “Device/Link Loads”• Represent percentage of capacity being used • Can be imported or manually created
• Flows • Visualized as demand objects• Can be imported or manually created
• Packets• Provide detailed performance statistics
• Delay, response time, jitter• Represent different types of application traffic• Can be imported from optional ACE module or manually created
• All three can exist simultaneously in the network model
129
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Importing and Characterizing Applications• With the Application Characterization Environment (ACE) module,
you can model any networked application by capturing the packets associated with that application, then importing them into Modeler.
130
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Link Loads
• Represent utilization of individual links
• Values are represented in percentage utilization for each time period• Utilization A->B
• Utilization B->A
• To set link loads manually:• Right-click a link, edit attributes
• Expand “Background Utilization” attribute
• Set utilization percentages for multiple time periods
• To import data from management platforms:• Traffic / Import Device/Link Loads
• MRTG
• InfoVista
• Concord eHealth-Network
• Text Files
131
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Flows
• Represent flow of network traffic between sources and destinations in network
• Flows are represented in bits/sec and packets/sec
• Flow data can be imported from management platforms• Cisco NetFlow (AS Aggregation or No Aggregation)
• Cflowd
• NetScout
• NetScout Ngenius
• NAI Distributed Sniffer / Sniffer Pro
• Agilent NetMetrix
• Spreadsheet / text files
• You can add new flows manually by choosing: Protocols / IP / Demands / Configure Traffic Demands Among Selected Nodes
132
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Flows
• Flows are a type of traffic demand. Demands appear as dotted blue lines from source to destination.
• Right-click a traffic demand to edit its attributes or hide the demand
• Use View / Demand Objects to hide/show all demands
• Use Traffic / Open Flows Browser to quickly examine many flows
133
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Packets
• Only used for discrete event simulation (DES) • Represent individual packets in network
• Application traffic• LSAs
• Application Configuration node used to globally define application traffic
• LSAs automatically sent by routers during simulation• Capture the transient behavior of the network
• Convergence• Latency• Queue depth • Protocol effects
• Verify Service Level Agreement compliance
134
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Converting Loads to Flows
• Import load traffic
• Gather information about traffic characteristics in real network
• Traffic / Convert Link Utilizations to Flows…
• Give Modeler constraints:
• List of possible endpoints of traffic flows
• Max/min amounts of traffic per endpoint
• Modeler performs conversion
• Caveat: This is an approximation!
• More constraints = more realistic flows
• Flow traffic can be used for failure, routing, and QoS studies
135
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Summary: Traffic Modeling
• Different traffic types available for different types of studies
• Three types available for import or creation:• Flows
• Loads
• Packets
• Flows are best choice for failure studies
• Loads are often easiest to obtain and best for viewing utilization of current network
• Packets are only used in discrete event simulation. Packets give the most detailed results and are the best choice for QoS studies.
136
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Object Reduction
• Each and every object in a model uses memory.
• Reducing the number of objects in a simulation allows for more efficient simulations, possibly allowing the workstation to overcome memory limitations (reduce swapping).
• Modeler provides LANs, Clouds and devices with varying interfaces to allow you to keep the number of objects in the network model to a minimum without sacrificing precision.
137
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Object Reduction
• LANs allow you to represent any number of workstations and a server as a single object.
• IP Clouds abstract numerous IP devices into a single cloud object.
• Each interface in a device requires memory. Creating devices that have the minimum amount of interfaces required for a model reduces memory needed to run simulation. • Example-- If all the hubs in a model will only have 8 connections, it is more
efficient to use an ethernet8_hub than an ethernet64_hub. Even if this doesn’t represent your actual hardware, it will not affect the precision of the simulation.
138
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Alternative Solutions
• There are other solutions that don’t involve model design.
• These alternative solutions can be characterized as “brute force” methods.
• The advantage to alternative solutions is that it gives you more latitude in your model designs.
• Increasing physical memory is always an effective way to increase workstation performance.
• Increasing the swap space is another effective method to get a better performance from a computer.
139
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Controlling Simulation Runtime
• Every protocol in Modeler has attributes associated with them. These attributes represent real-world application of a protocols procedures and techniques.
• In many cases, although representing a protocol’s actual behavior, certain attributes may not add value to a user’s model. In fact they will generate events that are not helpful to the results of the model and will actually increase simulation run-time.
• Modeler accounts for this by enabling certain efficiency modes in order to decrease the number of unimportant events and accelerate simulation run-time.
140
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Controlling Simulation Runtime
• When configuring a simulation, you will be able to view and edit a list of simulation attributes. These attributes will change based on the protocols being modeled.
• Among these attributes are a variety of efficiency modes that are enabled.
• Right-clicking on a simulation attribute will provide you with a complete description of that attribute.
141
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Importing Topology Information
• OPNET can import topologies from• ACE
• ATM Text files
• HP Network Node Manager
• Device Configurations (MVI Module)
• XML files
• VNE Server
142
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Aliases• An alias is a name used to reference a particular node in OPNET.
• An alias may be an IP address, MAC address or a user friendly name.
• An object may have multiple aliases.
• These aliases are used as endpoints for imported traffic.
• LAN models retain the aliases for all nodes that are contained in the LAN.
• This makes it possible to address traffic to the appropriate LAN model.
143
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Navigating Large Topologies
• OPNET provides you with various tools to facilitate the locating and selecting of nodes in a large network:
• Select Objects Logically
• Network Browser
144
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Edit / Select Objects
• Choose Edit / Select Objects to select certain objects throughout network based on the value of attributes.
• Ex: Select all links with a data rate greater than 1.544 Mbps
145
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Network Browser
• The Network Browser can be used to locate objects in your network.
• Choose View / Show Network Browser
• Type a search string into the “Find” box to show objects with that string in their names.
146
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Network Browser• To view an object’s attributes, right-click the object in the workspace
or the list and select “Edit Attributes.”
147
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Managing Projects and Scenarios
• Managing scenarios allows you to examine the status of scenario results, change project/scenario names and run multiple simulations sequentially.
Change the scenario name here.
Change the simulation duration.
Collect results on multiple scenarios by changing the results column to <recollect>. Clicking OK will start a simulation run for each scenario with <recollect> set.
Change the project name here.
Ok will write all of the changes to the project.
148
Modeler Day 3
© copyright 2003 OPNET Technologies, Inc.
Summary• Networks can be imported in a variety of manners.
• Modeler’s power as modeling software becomes evident with the analysis of large networks. As the size of the model increases so does the importance of efficient modeling.
• Avoiding common pitfalls and understanding the efficiency techniques inherent in Modeler leads to useful results and efficient models.
• By skillfully using the combination of the three types of traffic, you can achieve a comfortable balance of performance and precision.