Capacity Planning
for Enterprise
Networks
An Overview of Monitoring,
Baselining, Traffic Analysis,
and Performance
Optimization
0972_03F8_c3
NW98_US_601 1
2
0972_03F8_c3
NW98_US_601
Jake Hartinger
jharting@cisco.com
3
0972_03F8_c3
NW98_US_601
Capacity Planning Outline
Introduction
Collection and reporting of capacity
information
WAN capacity planning
LAN capacity planning
Upgrades and service level agreements
4
0972_03F8_c3
NW98_US_601
Increasing Importance
of Capacity Planning
85% of new application deployments fail to
meet service level agreements
Increased network complexity
High bandwidth multimedia applications and
increased burstiness
Increased reliance on network services for
business applications
Increased use of the network for
non-business activities
More organizations upgrading existing
networks and applications rather than
building new networks
5
0972_03F8_c3
NW98_US_601
Effective Capacity
Management
Performance
Performance
Baselining
Baselining
Observe Statistics
Observe Statistics
Collect Capacity
Collect Capacity
Data
Data
Analyze Traffic
Analyze Traffic
Solve Problems
Solve Problems
Plan Changes
Plan Changes
Evaluate
Evaluate
What-if
What-if
Analysis
Analysis
Gather
Configuration
and Traffic
Information
Implement
Changes
6
0972_03F8_c3
NW98_US_601
Resource Constraints
or Bottlenecks
Network or End Device CPU
Network or End Device I/O,
forwarding rates and latency
Network or End Device memory
and buffering
Interface and pipe sizes
Speed and distance
Application characteristics
7
0972_03F8_c3
NW98_US_601
Baseline Network
Performance and Capacity
Interface utilization
Device CPU, memory,
buffer,
I/O utilization
Network overhead
Raw performance
characteristics
Monthly or quarterly
baseline report
Performance
Performance
Baselining
Baselining
Observe Statistics
Collect Capacity
Data
Analyze Traffic
Solve Problems
Plan Changes
Evaluate
What-if
Analysis
Gather
Configuration
and Traffic
Information
Implement
Changes
8
0972_03F8_c3
NW98_US_601
Baseline Application
Performance and Capacity
Important for new or
customer funded
applications
Determine
application
behavior and
data flows
Rank mission critical
applications
Compare against
network capacity
Performance
Performance
Baselining
Baselining
Observe Statistics
Collect Capacity
Data
Analyze Traffic
Solve Problems
Plan Changes
Evaluate
What-if
Analysis
Gather
Configuration
and Traffic
Information
Implement
Changes
9
0972_03F8_c3
NW98_US_601
What-if Scenarios
Simulation applications
Lab application
modeling
Protocol Analyzer,
WAN emulator,
packet generator,
NETSYS
performance analyzer
Lab network modeling
NETSYS performance analyzer,
lab network modeling
Performance
Baselining
Observe Statistics
Collect Capacity
Data
Analyze Traffic
Solve Problems
Plan Changes
Evaluate
What-if
What-if
Analysis
Analysis
Gather
Configuration
and Traffic
Information
Implement
Changes
10
0972_03F8_c3
NW98_US_601
Collecting and Reporting
Capacity Information
Development of
information
collection plan
Tools for collecting
capacity
information
Defining capacity areas
Reporting and
interpreting results
Performance
Baselining
Observe Statistics
Observe Statistics
Collect Capacity
Collect Capacity
Data
Data
Analyze Traffic
Analyze Traffic
Solve Problems
Plan Changes
Evaluate
What-if
Analysis
Gather
Configuration
and Traffic
Information
Implement
Changes
11
0972_03F8_c3
NW98_US_601
Information to Collect
Link utilization
CPU, process CPU
Device throughput
Memory
Log file messages
Performance
(ping response time)
Queue drops
ATM parameters
Frame Relay DE,
FECN, BECN,
traffic-shaping
parameters
NetFlow and IP
accounting
Application
characteristics
12
0972_03F8_c3
NW98_US_601
Useful for modeling, design strategies,
capacity reporting, setting thresholds,
and service level agreements
Examples of capacity areas
LAN, core, distribution, and user
WAN, critical WAN, and individual WAN link
Access, ISDN, or dial
Network “class of service”
Application “class of service”
Defining Capacity Areas
13
0972_03F8_c3
NW98_US_601
Device thresholds exceeded by
capacity area
Pipe size thresholds exceeded by
capacity area
Network availability metrics by
capacity area
Threshold analysis
Why was threshold exceeded?
Trends and SLA agreements
Recommendations and
network changes
Network “Class of Service”
14
0972_03F8_c3
NW98_US_601
Application model, “what-if” scenario
Application performance thresholds
exceeded by capacity area
Application threshold analysis
Why was threshold exceeded?
Trends and SLA agreements
Recommendations and
network changes
Application “Class of Service”
15
0972_03F8_c3
NW98_US_601
Capacity Plan Strategy
Information to collect
Application and platform
Archival period, (database capacity)
Collection intervals
Reporting mechanisms
Monthly or quarterly capacity report
Upgrade criteria
Budget
Resource or outsource?
16
0972_03F8_c3
NW98_US_601
Tools for Capacity Planning
NETSYS connectivity baseliner
and solver
NETSYS performance baseliner
and solver
SwitchProbe®
products
TrafficDirector™
FlowCollector and FlowAnalyzer
17
0972_03F8_c3
NW98_US_601
NETSYS Connectivity
Baseliner and Solver
Test end-to-end reachability
Validate access list
Determine routing loops
Evaluate routing
protocol migration
Analyze and troubleshoot
connectivity configuration
Conduct what-if analysis of
changes and failure
18
0972_03F8_c3
NW98_US_601
NETSYS Performance
Baseliner and Solver
Analyze interaction
between application
traffic and network
configuration
Integrated data
acquisition and analysis
Determine resource
efficiency and end-to-
end performance analysis
Conduct what-if analysis
of optimization changes
Add on to connectivity tools
19
0972_03F8_c3
NW98_US_601
NETSYS Advisor
Automatic data collection
Maintains regularly updated status of network
configurations, traffic and performance data,
and observed routing tables
Automatic reporting, web based
Gives operations staff and managers greater control
of router configurations and end-to-
end network performance
Automatic diagnosis
Solve difficult routing problems
HTML/Java interface
Any user, any place, any time
Add on to connectivity tools
20
0972_03F8_c3
NW98_US_601
Family of standalone enhanced RMON
probes to monitor Fast Ethernet,
FDDI, Token Ring,
and Ethernet
Connect to critical interswitch and
server links and to Port Analyzer,
(SPAN), ports to provide full visibility
of a switched internetwork
Cisco SwitchProbe Products
21
0972_03F8_c3
NW98_US_601
Graphical user interface (GUI) based
RMON console manager
Provides extensive graphing, alarm,
logging, and reporting capabilities
TrafficDirector
22
0972_03F8_c3
NW98_US_601
Flow Analysis
FlowCollector
Collect and manage network flows
FlowAnalyzer
Display, monitor, and analyze network flows
Requires IOS 11.2 flow switching
23
0972_03F8_c3
NW98_US_601
Reporting and
Interpreting Results
Interpreting SNMP utilization levels
Understanding vendor thresholds
Setting exception levels
24
0972_03F8_c3
NW98_US_601
Peak and Average
Utilization Issues
Solution to narrow collection
interval
Low collection interval = high
overhead
Recommend >=5 minutes
Peak values not quite what
they seem
Close to threshold indicates
likely
exceed condition
Time in Collection
Intervals
Capacity
Parameter
(Threshold)
Green Bar Represents
SNMP Value Returned
25
0972_03F8_c3
NW98_US_601
Overhead Utilization
All non-user traffic
SNMP
Level 3 routing updates
CDP
Level 2 signaling
Other broadcast/multicast
Lowest level of traffic seen on link
Often 10% of link capacity or 20% of CPU
Important to understand and measure
in baselining
26
0972_03F8_c3
NW98_US_601
Often less than media speed for one
or more interfaces in network device
May be CPU, bus, I/O, memory, buffer,
or interface limitations
Work with vendor to understand
limitations, threshold behavior,
and tuning capabilities
Use vendor thresholds to reset overall
threshold levels for capacity reporting
Understanding Vendor
Thresholds
27
0972_03F8_c3
NW98_US_601
WAN Capacity Planning
Network and application
capacity issues
Building manageable WAN
solutions
WAN optimization
Device and
bandwidth thresholds
Developing manager
and user acceptance,
service level agreement,
and upgrade plan
Performance
Baselining
Observe Statistics
Collect Capacity
Data
Analyze Traffic
Solve Problems
Solve Problems
Plan Changes
Plan Changes
Evaluate
Evaluate
What-if
Analysis
Gather
Configuration
and Traffic
Information
Implement
Changes
28
0972_03F8_c3
NW98_US_601
Network and Application
Capacity Issues
Determining application
and network bandwidth
requirements
Application packet size
affect on performance
Application data flows
affect on performance
Batch requirements
Burst requirements
Latency issues
Speed and bandwidth
Network congestion
behavior
Variable performance
tolerance
Application prioritization
Affect of broadcast or
multicast in NBMA
environment
Behavior of UDP vs. TCP
applications
29
0972_03F8_c3
NW98_US_601
Speed and Bandwidth
Ping from San Francisco
to New York = 100ms
Ping from
San Francisco to
Los Angeles = 45ms
Ping from San Francisco
to Los Angeles with
frame congestion = 100ms
Los Angeles
New York
San Francisco
T-1
56kbps/
16kbps CIR
Frame Relay
Cloud
30
0972_03F8_c3
NW98_US_601
WAN Connectivity Options
Leased Line
Leased Line Frame Relay
Frame Relay ATM
ATM
56kbps and Above
56kbps and Above
Optimizes Speed
Optimizes Speed
Minimizes Variable
Minimizes Variable
Performance
Performance
Easier
Easier
Management
Management
Highest Cost
Highest Cost
56kbps to
56kbps to
768kbps
768kbps
Variable
Variable
Performance
Performance
More
More
Performance
Performance
Management
Management
Lowest Cost
Lowest Cost
T-1 and Above
T-1 and Above
Bandwidth
Bandwidth
Flexibility
Flexibility
Cell Overhead
Cell Overhead
Cloud
Cloud
Redundancy
Redundancy
Highest
Highest
Bandwidth/$
Bandwidth/$
31
0972_03F8_c3
NW98_US_601
Frame Relay
Subscription Example
Boston
256 Kbps Port
Dallas
128 Kbps Port
Atlanta
128 Kbps Port
New York
512 Kbps Port
Host
Host
New York to Boston PVC 128 Kbps CIR
Boston to New York PVC 32 Kbps CIR
New York to Dallas 64 Kbps CIR
Dallas to New York 32 Kbps CIR
New York to Atlanta 96 Kbps CIR
Atlanta to New York 32 Kbps CIR
32
0972_03F8_c3
NW98_US_601
ATM Subscription Example
Carrier Cloud
PBX
PBX
PBX
1
1 3
3
1
1
2
2 1
1
3
3
UBR Virtual Path
CBR Virtual Circuit
UBR Virtual Circuit
VBR-RT Virtual Circuit
33
0972_03F8_c3
NW98_US_601
WAN Optimization
Queuing methodologies
FIFO queuing, weighted-fair queuing,
custom queuing, priority queuing,
split-DLCI queuing
Traffic shaping
Frame Relay traffic shaping, generic traffic
shaping, ATM traffic shaping, IP precedence,
RED, WRED, CAR
Compression
Link overhead
34
0972_03F8_c3
NW98_US_601
Queuing’s Premise
Getting better service is a matter
of managing congested queues
Over-all latency and bandwidth
are constant
Make some traffic absorb latency,
and therefore give up bandwidth
Shield other traffic from latency,
and therefore gain bandwidth
35
0972_03F8_c3
NW98_US_601
Queuing Comparison
FIFO queuing
Least resources, simple
Unpredictable QoS
Weighted-fair queuing
Predictable performance
May be “unfair”,
more sorting
Priority queuing
Based on access-list class
Unpredictable performance for
queue congestion, can cause
traffic lockout
Custom queuing
Base queuing on
fiscal input
Unpredictable in class
Split DLCI priority
Solves Frame Relay
congestion problem
Unpredictable within queue,
traffic lockout
36
0972_03F8_c3
NW98_US_601
FIFO Queuing Affect
on Performance
Milliseconds
0
500
1000
1500
2000
2500
3000
3500
0 50 100 150 200 250 300 350 400 450 500 550 600
File Transfer
Application X
Ping
37
0972_03F8_c3
NW98_US_601
Weighted-Fair Queuing
Affect on Performance
File Transfer
Application X
Ping
Milliseconds
0
50
100
150
200
250
300
0 50 100 150 200 250 300 350 400 450 500 550 600
38
0972_03F8_c3
NW98_US_601
Traffic Shaping
ATM traffic shaping
Traffic classes, CBR, VBR-RT, VBR, ABR
Service variables: PCR, CDV, CLR, MCR
Leaky bucket—flow control algorithm
whereby a “bucket” receives credits at
a rate equal to a user’s traffic-shaping
parameters, overtime credits used at
average rate but allows for bursts
39
0972_03F8_c3
NW98_US_601
Traffic Shaping
Frame Relay traffic shaping
Specified by mean-rate (normally CIR), burst
rate, (Bc), and excess burst, (Be)
Used to prevent loss in Frame Relay cloud and
FECN/BECN reaction to improve
Frame Relay performance
Generic traffic shaping
Specified by mean-rate, burst-size, and excess
burst size
Used by service providers to subrate
traffic on interfaces < E1
40
0972_03F8_c3
NW98_US_601
Traffic Shaping
IP precedence
Prioritization within queuing mechanism using
3 bit IP packet precedence field
RED
Congestion avoidance for TCP applications;
WRED based on IP precedence
41
0972_03F8_c3
NW98_US_601
Traffic Shaping
CAR
Allows network administrator to classify traffic
classify traffic
in order to rate limit
rate limit it based on a
token bucket scheme
RSVP
Reservation protocol for multimedia
applications to guarantee bandwidth
Will help overcome jitter, delay, loss,
insufficient bandwidth
42
0972_03F8_c3
NW98_US_601
Compression and Link Overhead
Compression
Hardware compression
Cisco CSA
External device
Software
compression
FRF.9
Cisco Frame Relay
Voice compression
32/24/16 ADPCM
16/8 CELP
Minimal link
overhead
All non-user traffic
SNMP
Level 3 routing updates
CDP
Level 2 signaling
Other broadcast
and multicast
43
0972_03F8_c3
NW98_US_601
WAN Service Level
Agreements and Upgrades
Baseline performance
Document consistent level of performance
by application
Number of users, transactions, performance
Access support
Create upgrade criteria based on upgrade
plan
New application requirements
Increasing users
Exception reporting
44
0972_03F8_c3
NW98_US_601
LAN Capacity Planning
Application and network
requirements
Building manageable LAN
solutions
LAN optimization
Device and
bandwidth thresholds
Developing manager, user
acceptance,
(service level agreement)
and upgrade plan
Performance
Baselining
Observe Statistics
Collect Capacity
Data
Analyze Traffic
Solve Problems
Solve Problems
Plan Changes
Plan Changes
Evaluate
Evaluate
What-if
Analysis
Gather
Configuration
and Traffic
Information
Implement
Changes
45
0972_03F8_c3
NW98_US_601
Application and Network
Requirements
Multicast and broadcast
Voice and video
File transfers and backups (server to server)
Centralized servers/workgroup servers
Routed and routing protocols
General data flows and volumes
Availability/redundancy
46
0972_03F8_c3
NW98_US_601
Hubs, Switches, and Routers
Hubs
10 or 100 Ethernet
100 FDDI
Inexpensive
No level II or III filter
Switches
10 or 100 Ethernet
ATM
Inexpensive
No multicast/
broadcast filter
High forwarding rates
Router
Any interface
Complete and flexible
level II, III filters
Rich features sets
Expensive
Forwarding rates lower
than switches
47
0972_03F8_c3
NW98_US_601
LAN Solutions: Station Layer
Switched vs. non-switched
Collision domain for Ethernet
Station bandwidth switched
vs. non-switched
Broadcast effect on overall
bandwidth and system CPU,
“background radiation”
Station bandwidth requirements
FDDI, FE, Token Ring, ATM
Expectation for Multicast
Fast-Ethernet Hub
FDDI/CDDI Hub
Workgroup Switch
Workgroup Switch
Workgroup Hub
48
0972_03F8_c3
NW98_US_601
LAN Solutions:
Distribution Layer
Protocols and
broadcasts, IP, IPX®
,
AppleTalk, Netbios
VLANs, RSM, and ISL
Uplink congestion point
to backbone
Distribution layer
servers
Load sharing,
redundancy, (HSRP)
49
0972_03F8_c3
NW98_US_601
Protocols and level III
capacity issues
Level III switching performance
Bandwidth
requirements
Hierarchy for capacity
Backbone layer servers
ATM backbones
NBMA issues, VC issues
Load sharing, redundancy
LAN Solutions: Backbone Layer
50
0972_03F8_c3
NW98_US_601
LAN Optimization
Bandwidth bottlenecks
Background radiation
Adding routing and switching
Multicast filtering, CGMP
Load sharing and redundancy
CPU and switching paths
Queues and buffers
Routing convergence
51
0972_03F8_c3
NW98_US_601
Baseline network capacity
SLA based on availability/performance
model
Create upgrade criteria based
on upgrade plan
Multicast or other application requirement
Increasing Enterprise, building/user model
Exception reporting
LAN Service Level
Agreements and Upgrades
52
0972_03F8_c3
NW98_US_601
Upgrades and Service
Level Agreements
Document expected
performance/capacity improvement
Confirm expectations after upgrade
Create new baseline measurements
Create new exception reporting
measurements
Update service level agreements
53
0972_03F8_c3
NW98_US_601
Service Level Agreement
Network service level agreement
Ping response and response time, capacity
Availability expectation by location or group
Performance expectation by location or group
Rules for non-conformance
Application service level agreement
Application response time, (application call or ping)
Network availability for application
Performance for application
Rules for non-conformance
Referenced document and
available metrics
54
0972_03F8_c3
NW98_US_601

Network-planning-Network and performance.ppt

  • 1.
    Capacity Planning for Enterprise Networks AnOverview of Monitoring, Baselining, Traffic Analysis, and Performance Optimization 0972_03F8_c3 NW98_US_601 1
  • 2.
  • 3.
    3 0972_03F8_c3 NW98_US_601 Capacity Planning Outline Introduction Collectionand reporting of capacity information WAN capacity planning LAN capacity planning Upgrades and service level agreements
  • 4.
    4 0972_03F8_c3 NW98_US_601 Increasing Importance of CapacityPlanning 85% of new application deployments fail to meet service level agreements Increased network complexity High bandwidth multimedia applications and increased burstiness Increased reliance on network services for business applications Increased use of the network for non-business activities More organizations upgrading existing networks and applications rather than building new networks
  • 5.
    5 0972_03F8_c3 NW98_US_601 Effective Capacity Management Performance Performance Baselining Baselining Observe Statistics ObserveStatistics Collect Capacity Collect Capacity Data Data Analyze Traffic Analyze Traffic Solve Problems Solve Problems Plan Changes Plan Changes Evaluate Evaluate What-if What-if Analysis Analysis Gather Configuration and Traffic Information Implement Changes
  • 6.
    6 0972_03F8_c3 NW98_US_601 Resource Constraints or Bottlenecks Networkor End Device CPU Network or End Device I/O, forwarding rates and latency Network or End Device memory and buffering Interface and pipe sizes Speed and distance Application characteristics
  • 7.
    7 0972_03F8_c3 NW98_US_601 Baseline Network Performance andCapacity Interface utilization Device CPU, memory, buffer, I/O utilization Network overhead Raw performance characteristics Monthly or quarterly baseline report Performance Performance Baselining Baselining Observe Statistics Collect Capacity Data Analyze Traffic Solve Problems Plan Changes Evaluate What-if Analysis Gather Configuration and Traffic Information Implement Changes
  • 8.
    8 0972_03F8_c3 NW98_US_601 Baseline Application Performance andCapacity Important for new or customer funded applications Determine application behavior and data flows Rank mission critical applications Compare against network capacity Performance Performance Baselining Baselining Observe Statistics Collect Capacity Data Analyze Traffic Solve Problems Plan Changes Evaluate What-if Analysis Gather Configuration and Traffic Information Implement Changes
  • 9.
    9 0972_03F8_c3 NW98_US_601 What-if Scenarios Simulation applications Labapplication modeling Protocol Analyzer, WAN emulator, packet generator, NETSYS performance analyzer Lab network modeling NETSYS performance analyzer, lab network modeling Performance Baselining Observe Statistics Collect Capacity Data Analyze Traffic Solve Problems Plan Changes Evaluate What-if What-if Analysis Analysis Gather Configuration and Traffic Information Implement Changes
  • 10.
    10 0972_03F8_c3 NW98_US_601 Collecting and Reporting CapacityInformation Development of information collection plan Tools for collecting capacity information Defining capacity areas Reporting and interpreting results Performance Baselining Observe Statistics Observe Statistics Collect Capacity Collect Capacity Data Data Analyze Traffic Analyze Traffic Solve Problems Plan Changes Evaluate What-if Analysis Gather Configuration and Traffic Information Implement Changes
  • 11.
    11 0972_03F8_c3 NW98_US_601 Information to Collect Linkutilization CPU, process CPU Device throughput Memory Log file messages Performance (ping response time) Queue drops ATM parameters Frame Relay DE, FECN, BECN, traffic-shaping parameters NetFlow and IP accounting Application characteristics
  • 12.
    12 0972_03F8_c3 NW98_US_601 Useful for modeling,design strategies, capacity reporting, setting thresholds, and service level agreements Examples of capacity areas LAN, core, distribution, and user WAN, critical WAN, and individual WAN link Access, ISDN, or dial Network “class of service” Application “class of service” Defining Capacity Areas
  • 13.
    13 0972_03F8_c3 NW98_US_601 Device thresholds exceededby capacity area Pipe size thresholds exceeded by capacity area Network availability metrics by capacity area Threshold analysis Why was threshold exceeded? Trends and SLA agreements Recommendations and network changes Network “Class of Service”
  • 14.
    14 0972_03F8_c3 NW98_US_601 Application model, “what-if”scenario Application performance thresholds exceeded by capacity area Application threshold analysis Why was threshold exceeded? Trends and SLA agreements Recommendations and network changes Application “Class of Service”
  • 15.
    15 0972_03F8_c3 NW98_US_601 Capacity Plan Strategy Informationto collect Application and platform Archival period, (database capacity) Collection intervals Reporting mechanisms Monthly or quarterly capacity report Upgrade criteria Budget Resource or outsource?
  • 16.
    16 0972_03F8_c3 NW98_US_601 Tools for CapacityPlanning NETSYS connectivity baseliner and solver NETSYS performance baseliner and solver SwitchProbe® products TrafficDirector™ FlowCollector and FlowAnalyzer
  • 17.
    17 0972_03F8_c3 NW98_US_601 NETSYS Connectivity Baseliner andSolver Test end-to-end reachability Validate access list Determine routing loops Evaluate routing protocol migration Analyze and troubleshoot connectivity configuration Conduct what-if analysis of changes and failure
  • 18.
    18 0972_03F8_c3 NW98_US_601 NETSYS Performance Baseliner andSolver Analyze interaction between application traffic and network configuration Integrated data acquisition and analysis Determine resource efficiency and end-to- end performance analysis Conduct what-if analysis of optimization changes Add on to connectivity tools
  • 19.
    19 0972_03F8_c3 NW98_US_601 NETSYS Advisor Automatic datacollection Maintains regularly updated status of network configurations, traffic and performance data, and observed routing tables Automatic reporting, web based Gives operations staff and managers greater control of router configurations and end-to- end network performance Automatic diagnosis Solve difficult routing problems HTML/Java interface Any user, any place, any time Add on to connectivity tools
  • 20.
    20 0972_03F8_c3 NW98_US_601 Family of standaloneenhanced RMON probes to monitor Fast Ethernet, FDDI, Token Ring, and Ethernet Connect to critical interswitch and server links and to Port Analyzer, (SPAN), ports to provide full visibility of a switched internetwork Cisco SwitchProbe Products
  • 21.
    21 0972_03F8_c3 NW98_US_601 Graphical user interface(GUI) based RMON console manager Provides extensive graphing, alarm, logging, and reporting capabilities TrafficDirector
  • 22.
    22 0972_03F8_c3 NW98_US_601 Flow Analysis FlowCollector Collect andmanage network flows FlowAnalyzer Display, monitor, and analyze network flows Requires IOS 11.2 flow switching
  • 23.
    23 0972_03F8_c3 NW98_US_601 Reporting and Interpreting Results InterpretingSNMP utilization levels Understanding vendor thresholds Setting exception levels
  • 24.
    24 0972_03F8_c3 NW98_US_601 Peak and Average UtilizationIssues Solution to narrow collection interval Low collection interval = high overhead Recommend >=5 minutes Peak values not quite what they seem Close to threshold indicates likely exceed condition Time in Collection Intervals Capacity Parameter (Threshold) Green Bar Represents SNMP Value Returned
  • 25.
    25 0972_03F8_c3 NW98_US_601 Overhead Utilization All non-usertraffic SNMP Level 3 routing updates CDP Level 2 signaling Other broadcast/multicast Lowest level of traffic seen on link Often 10% of link capacity or 20% of CPU Important to understand and measure in baselining
  • 26.
    26 0972_03F8_c3 NW98_US_601 Often less thanmedia speed for one or more interfaces in network device May be CPU, bus, I/O, memory, buffer, or interface limitations Work with vendor to understand limitations, threshold behavior, and tuning capabilities Use vendor thresholds to reset overall threshold levels for capacity reporting Understanding Vendor Thresholds
  • 27.
    27 0972_03F8_c3 NW98_US_601 WAN Capacity Planning Networkand application capacity issues Building manageable WAN solutions WAN optimization Device and bandwidth thresholds Developing manager and user acceptance, service level agreement, and upgrade plan Performance Baselining Observe Statistics Collect Capacity Data Analyze Traffic Solve Problems Solve Problems Plan Changes Plan Changes Evaluate Evaluate What-if Analysis Gather Configuration and Traffic Information Implement Changes
  • 28.
    28 0972_03F8_c3 NW98_US_601 Network and Application CapacityIssues Determining application and network bandwidth requirements Application packet size affect on performance Application data flows affect on performance Batch requirements Burst requirements Latency issues Speed and bandwidth Network congestion behavior Variable performance tolerance Application prioritization Affect of broadcast or multicast in NBMA environment Behavior of UDP vs. TCP applications
  • 29.
    29 0972_03F8_c3 NW98_US_601 Speed and Bandwidth Pingfrom San Francisco to New York = 100ms Ping from San Francisco to Los Angeles = 45ms Ping from San Francisco to Los Angeles with frame congestion = 100ms Los Angeles New York San Francisco T-1 56kbps/ 16kbps CIR Frame Relay Cloud
  • 30.
    30 0972_03F8_c3 NW98_US_601 WAN Connectivity Options LeasedLine Leased Line Frame Relay Frame Relay ATM ATM 56kbps and Above 56kbps and Above Optimizes Speed Optimizes Speed Minimizes Variable Minimizes Variable Performance Performance Easier Easier Management Management Highest Cost Highest Cost 56kbps to 56kbps to 768kbps 768kbps Variable Variable Performance Performance More More Performance Performance Management Management Lowest Cost Lowest Cost T-1 and Above T-1 and Above Bandwidth Bandwidth Flexibility Flexibility Cell Overhead Cell Overhead Cloud Cloud Redundancy Redundancy Highest Highest Bandwidth/$ Bandwidth/$
  • 31.
    31 0972_03F8_c3 NW98_US_601 Frame Relay Subscription Example Boston 256Kbps Port Dallas 128 Kbps Port Atlanta 128 Kbps Port New York 512 Kbps Port Host Host New York to Boston PVC 128 Kbps CIR Boston to New York PVC 32 Kbps CIR New York to Dallas 64 Kbps CIR Dallas to New York 32 Kbps CIR New York to Atlanta 96 Kbps CIR Atlanta to New York 32 Kbps CIR
  • 32.
    32 0972_03F8_c3 NW98_US_601 ATM Subscription Example CarrierCloud PBX PBX PBX 1 1 3 3 1 1 2 2 1 1 3 3 UBR Virtual Path CBR Virtual Circuit UBR Virtual Circuit VBR-RT Virtual Circuit
  • 33.
    33 0972_03F8_c3 NW98_US_601 WAN Optimization Queuing methodologies FIFOqueuing, weighted-fair queuing, custom queuing, priority queuing, split-DLCI queuing Traffic shaping Frame Relay traffic shaping, generic traffic shaping, ATM traffic shaping, IP precedence, RED, WRED, CAR Compression Link overhead
  • 34.
    34 0972_03F8_c3 NW98_US_601 Queuing’s Premise Getting betterservice is a matter of managing congested queues Over-all latency and bandwidth are constant Make some traffic absorb latency, and therefore give up bandwidth Shield other traffic from latency, and therefore gain bandwidth
  • 35.
    35 0972_03F8_c3 NW98_US_601 Queuing Comparison FIFO queuing Leastresources, simple Unpredictable QoS Weighted-fair queuing Predictable performance May be “unfair”, more sorting Priority queuing Based on access-list class Unpredictable performance for queue congestion, can cause traffic lockout Custom queuing Base queuing on fiscal input Unpredictable in class Split DLCI priority Solves Frame Relay congestion problem Unpredictable within queue, traffic lockout
  • 36.
    36 0972_03F8_c3 NW98_US_601 FIFO Queuing Affect onPerformance Milliseconds 0 500 1000 1500 2000 2500 3000 3500 0 50 100 150 200 250 300 350 400 450 500 550 600 File Transfer Application X Ping
  • 37.
    37 0972_03F8_c3 NW98_US_601 Weighted-Fair Queuing Affect onPerformance File Transfer Application X Ping Milliseconds 0 50 100 150 200 250 300 0 50 100 150 200 250 300 350 400 450 500 550 600
  • 38.
    38 0972_03F8_c3 NW98_US_601 Traffic Shaping ATM trafficshaping Traffic classes, CBR, VBR-RT, VBR, ABR Service variables: PCR, CDV, CLR, MCR Leaky bucket—flow control algorithm whereby a “bucket” receives credits at a rate equal to a user’s traffic-shaping parameters, overtime credits used at average rate but allows for bursts
  • 39.
    39 0972_03F8_c3 NW98_US_601 Traffic Shaping Frame Relaytraffic shaping Specified by mean-rate (normally CIR), burst rate, (Bc), and excess burst, (Be) Used to prevent loss in Frame Relay cloud and FECN/BECN reaction to improve Frame Relay performance Generic traffic shaping Specified by mean-rate, burst-size, and excess burst size Used by service providers to subrate traffic on interfaces < E1
  • 40.
    40 0972_03F8_c3 NW98_US_601 Traffic Shaping IP precedence Prioritizationwithin queuing mechanism using 3 bit IP packet precedence field RED Congestion avoidance for TCP applications; WRED based on IP precedence
  • 41.
    41 0972_03F8_c3 NW98_US_601 Traffic Shaping CAR Allows networkadministrator to classify traffic classify traffic in order to rate limit rate limit it based on a token bucket scheme RSVP Reservation protocol for multimedia applications to guarantee bandwidth Will help overcome jitter, delay, loss, insufficient bandwidth
  • 42.
    42 0972_03F8_c3 NW98_US_601 Compression and LinkOverhead Compression Hardware compression Cisco CSA External device Software compression FRF.9 Cisco Frame Relay Voice compression 32/24/16 ADPCM 16/8 CELP Minimal link overhead All non-user traffic SNMP Level 3 routing updates CDP Level 2 signaling Other broadcast and multicast
  • 43.
    43 0972_03F8_c3 NW98_US_601 WAN Service Level Agreementsand Upgrades Baseline performance Document consistent level of performance by application Number of users, transactions, performance Access support Create upgrade criteria based on upgrade plan New application requirements Increasing users Exception reporting
  • 44.
    44 0972_03F8_c3 NW98_US_601 LAN Capacity Planning Applicationand network requirements Building manageable LAN solutions LAN optimization Device and bandwidth thresholds Developing manager, user acceptance, (service level agreement) and upgrade plan Performance Baselining Observe Statistics Collect Capacity Data Analyze Traffic Solve Problems Solve Problems Plan Changes Plan Changes Evaluate Evaluate What-if Analysis Gather Configuration and Traffic Information Implement Changes
  • 45.
    45 0972_03F8_c3 NW98_US_601 Application and Network Requirements Multicastand broadcast Voice and video File transfers and backups (server to server) Centralized servers/workgroup servers Routed and routing protocols General data flows and volumes Availability/redundancy
  • 46.
    46 0972_03F8_c3 NW98_US_601 Hubs, Switches, andRouters Hubs 10 or 100 Ethernet 100 FDDI Inexpensive No level II or III filter Switches 10 or 100 Ethernet ATM Inexpensive No multicast/ broadcast filter High forwarding rates Router Any interface Complete and flexible level II, III filters Rich features sets Expensive Forwarding rates lower than switches
  • 47.
    47 0972_03F8_c3 NW98_US_601 LAN Solutions: StationLayer Switched vs. non-switched Collision domain for Ethernet Station bandwidth switched vs. non-switched Broadcast effect on overall bandwidth and system CPU, “background radiation” Station bandwidth requirements FDDI, FE, Token Ring, ATM Expectation for Multicast Fast-Ethernet Hub FDDI/CDDI Hub Workgroup Switch Workgroup Switch Workgroup Hub
  • 48.
    48 0972_03F8_c3 NW98_US_601 LAN Solutions: Distribution Layer Protocolsand broadcasts, IP, IPX® , AppleTalk, Netbios VLANs, RSM, and ISL Uplink congestion point to backbone Distribution layer servers Load sharing, redundancy, (HSRP)
  • 49.
    49 0972_03F8_c3 NW98_US_601 Protocols and levelIII capacity issues Level III switching performance Bandwidth requirements Hierarchy for capacity Backbone layer servers ATM backbones NBMA issues, VC issues Load sharing, redundancy LAN Solutions: Backbone Layer
  • 50.
    50 0972_03F8_c3 NW98_US_601 LAN Optimization Bandwidth bottlenecks Backgroundradiation Adding routing and switching Multicast filtering, CGMP Load sharing and redundancy CPU and switching paths Queues and buffers Routing convergence
  • 51.
    51 0972_03F8_c3 NW98_US_601 Baseline network capacity SLAbased on availability/performance model Create upgrade criteria based on upgrade plan Multicast or other application requirement Increasing Enterprise, building/user model Exception reporting LAN Service Level Agreements and Upgrades
  • 52.
    52 0972_03F8_c3 NW98_US_601 Upgrades and Service LevelAgreements Document expected performance/capacity improvement Confirm expectations after upgrade Create new baseline measurements Create new exception reporting measurements Update service level agreements
  • 53.
    53 0972_03F8_c3 NW98_US_601 Service Level Agreement Networkservice level agreement Ping response and response time, capacity Availability expectation by location or group Performance expectation by location or group Rules for non-conformance Application service level agreement Application response time, (application call or ping) Network availability for application Performance for application Rules for non-conformance Referenced document and available metrics
  • 54.

Editor's Notes

  • #1 3 3
  • #3 2 Introduction discusses the basic workflows that should be happening for quality capacity and performance planning. Collecting & Reporting This discusses what information needs to be collected and how to use it. The tool section simply gives a quick overview of Cisco tools that can help collect and report the required info. WAN Capacity Planning discusses specific issues in WAN Capacity planning and tools/commands that can help optimize capacity and performance. LAN Capacity Planning discusses specific issues in LAN capacity planning and tools/commands that can help optimize capacity and performance. A discussion of analog capacity planning is not included. Upgrades and Service Level agreements is the last section discussing how to provide levels of performance that meet organization requirements.
  • #4 3 85% figure taken from www.optimal.com. Optimal supplies Performance & Capacity applications that may compete with NETSYS performance analyzer.
  • #5 4 This is the workflow of correct capacity planning.
  • #6 5 These are the only resource restraints in a network. Some of this information can be collected via SNMP MIB variables. Other, including speed and application characteristics are more independent and require individual analysis.
  • #7 6 Baselining is perhaps the key to Capacity Planning. Baselining allows network administrators to understand their current level of capacity and performance which is needed to understand new network or application additions or to initially find potential problems. Baselining should be done with whatever information is available on a monthly or quarterly basis in the form of a baseline or capacity report which also helps identify trends.
  • #8 7 Baselining application characteristics is the key to successful application growth and performance, especially in WAN environments. Baselining new applications will determine whether the application deployment will be successful or whether additional resources are needed. Ranking applications is important. This should be done by the business itself so the network administrators can supply increasing quality of service for more important applications.
  • #9 8 What-if scenarios are done in three basic ways. one, with network simulation application that can understand performance capacity of your devices. Two, by physically modeling the application in a lab environment to understand exact resource and performance requirements. Three, by adding load in a lab test environment to understand the affect of additional load on the network. All are important however nothing can replace actually modeling an application or environment to understand the affects.
  • #10 9 This section should help organizations start planning what they are doing and understand what information they are currently collecting that might help baseline network & application capacity.
  • #11 10 examples of information that might be collected. Link utilization, CPU, log file messages and performance are critical.
  • #12 11 Breaking up capacity areas helps because they may have different requirements or even applications that are organization specific. Breaking into areas also allows network administrators to get a handle on capacity planning.
  • #13 12 Capacity Planning is done on two levels, Network and Application. Looking broader you might also add server and client however this is well beyond network capacity planning. Network defines the overall requirements. Application addresses individual application issues and requirements.
  • #14 13
  • #15 14 This slide is a simple overview on developing a capacity plan. It is assumed that most organization do little or no capacity planning. Capacity Planning is frequently forgotten because it is less reactive and requires proactive planning including budget, resources and planning.
  • #16 15 Cisco Tools for Capacity Planning. NETSYS also allows simulation modeling and what-if analysis.
  • #17 16 Connectivity tool to understand configuration errors and overall connectivity as derived from routing protocols. Helps to determine routing behavior when access-lsits and routing changes are simulated.
  • #18 17 This tool can help determine capacity exception areas and perform what-if capacity analysis based on added load or network changes.
  • #19 18 NETSYS advisor helps report performance issues routing issues.
  • #20 19 collects RMON information from a network segment useful in determining capacity and trending.
  • #21 20 SNMP management station useful to determine Capacity and utilization mostly in LAN environment.
  • #22 21 Flows help determine traffic and even individual application behavior and bandwidth. Flow analysis can also determine application volumes, top talkers and packet sizes.
  • #23 22 Preview of coming slides.
  • #24 23 Simple slide to show that SNMP utilization doesn't really collect Peak utilization.
  • #25 24 Overhead utilization useful for baselining and understanding non-user network overhead.
  • #26 25 Vendors have published thresholds for Capacity that may or may not be useful as numbers may be for competitive needs. Users should still understand capacity issues with their traffic, environment and mix of features.
  • #27 26 Issues tending to be more specific to WAN environments such as speed/distance and bandwidth.
  • #28 27 More detailed analysis may be required for WANs as bandwidth is expensive and chance of application deployment failure is much higher than in LAN environments.
  • #29 28 speed NOT EQUAL to bandwidth. Industry & Applications designers sometimes believe bigger pipes are faster just because they are bigger.
  • #30 29 Basic connectivity options. X.25 or dialup not used much in U.S. SMDS is somewhat rare.
  • #31 30 Example showing the potential Frame Relay subscription values. These are bandwidth and subscription level of main trunk bandwidth and subscription level of remote trunk CIR to remote CIR from remote
  • #32 31 Example subscription levels in ATM Carrier service network including service type: Virtual Path, Virtual Circuit, UBR, VBR, CBR service. Also PCR, MCR, CLR as configured in Carrier network. (Carriers define these differently!) and offer slightly different services.
  • #33 32 Overview of slides to come on basic performance and bandwidth optimization techniques.
  • #34 33 Basic idea of non FIFO queuing techniques.
  • #35 34 Split DLCI priority queuing also called “DLCI prioritization”.
  • #36 35 This is an example showing the affect when FIFO and Fair-Weighted Queuing are used when bandwidth contention is seen with 3 basic applications, FTP, application, (like SQLNET), and Ping.
  • #37 36
  • #38 37 Understanding of ATM traffic shaping capabilities and how traffic is being shaped in Carrier connection. Different Traffic classes have different mechanisms or values for PCR, CDV, CLR and MCR. Leaky Bucket - Flow control algorithm whereby a "bucket" receives credits at a rate equal to a user's traffic-shaping parameters, Overtime credits used at avg. rate but allows for bursts
  • #39 38 Basic concepts of Frame Relay and Generic Traffic Shaping.
  • #40 39 Basic concepts of IP Precedence and RED.
  • #41 40 Basic concept of CAR and RSVP.
  • #42 41 Basic concepts of compression and minimizing link overhead.
  • #43 42 This is an important step in Capacity Planning to become proactive rather than reactive to problems.
  • #44 43
  • #45 44 LAN Capacity planning may be more general, not requiring quite the application level of analysis that WAN Capacity Planning requires.
  • #46 45 Different LAN devices and affects on performance/capacity.
  • #47 46 LAN broken down into three areas of Capacity, Station, Distribution and backbone. This may be different for different environments, but overall, most all of the concepts will be important in LAN environments.
  • #48 47 Distribution Layer issues the same regardless of topology i.e. VLANs, RSM, ISL. The basic concepts are congestion points, overall switching capability, broadcast domains for different applications and protocols. Rules of thumb. IP=1000 or less in broadcast domain IPX=500 or less in broadcast domain AppleTalk=200 or less in broadcast domain.
  • #49 48 Important is the necessary level of hierarchy given the specific environment and feature set. Also covers ATM backbone issues, overall bandwidth/backplane requirements.
  • #50 49 Methods of improving or optimizing LAN environments. Should have capacity to handle basic LAN traffic.
  • #51 50 Important to be proactive rather than reactive to problems.
  • #52 51 emphasizing overall that upgrade plans, service level agreements allow network administrators to be proactive rather than reactive.
  • #53 52 Documents should be in writing. This provides ammunition to network administrators to make changes, sometimes costly when SLAs are not being met.