Requirements and Capacity Planning

1
Requirements and Capacity Planning

• A system approach to network design
• The analysis and design process
• User requirement
• Capacity planning
• Traffic characteristics
• Performance characteristics
• Systems and network services
• Characterizing services

2
The Analysis and Design Process

• Combination of several factors
• Design goals
• minimizing cost
• maximizing performance
• Trade-offs
• Cost versus performance
• Simplicity versus function
• Balance between architecture and function
• Hierarchy
• Redundancy (trade-off between these too)

3
The Analysis and Design Process

• Analysis
• Requirements analysis
• Flow analysis
• Design
• Logical design
• Physical design
• Addressing and routing not covered
• Implementation not covered
• Management not covered

4
Components of the Analysis and Design Process
Not covered
5
The Analysis and Design Process

• Identify the services and performance levels that
  the design must satisfy
• Choose the best network technologies and services
  to meet the design goals

6
Network services and Services-based Networking

• Levels of performance
• Round-trip delay
• Bandwidth
• Reliability
• Availability
• Functions, e.g.,
• Security levels
• User groups for virtual networks

7
Services-based networking
8
Systems and Network Services

• Systems
• Set of components that work together to support
  or provide
• connectivity,
• communications, and
• network services to users
• Generic components
• Users
• Applications
• Hosts and
• Networks

9
Systems and Network Services

• Users
• Network and computer support
• Developers
• Customers
• Applications
• Specific to a particular user/customer
• Generic to a customer base
• Generic to the entire network
• Hosts
• Subdivided to show the OS, device drivers, API
• Networks

10
Network Services

• Levels of performance and function offered by
  network/host/application
• Levels of performance capacity, delay,
  reliability
• Functions security, accounting, billing,
  scheduling, management
• Service characteristics
• The service offering and service request
• Important in configuring services in network
  elements (routers, switches, host OSs etc.)
• Used to configure, monitor, and verify services
  in the network

11
Network Services

• Service levels
• Service requirements or characteristics are
  grouped together to describe service levels for
  the network
• One can configure, measure and verify a service
  level instead of a number of service
  characteristics
• Also useful in accounting and billing

12
Network Services

• Describing service levels
• Committed information rates (CIRs)
• a specified amount of guaranteed bandwidth
  (measured in bits per second) - data rate not
  exceeding this level will be delivered
• Classes of service (CoS) delivery priority
• Types of service (ToS) e.g., minimize delay or
  maximize throughput
• Quality/(ies) of Service (QoS) in terms of delay

13
Characterizing Services

• Service requests
• Best effort
• Specified (deterministic and guaranteed) services
• Service offerings
• To meet requests (e.g., 1.5 Mbps, end-to-end
  delay lt40ms)
• ATM QoS
• Switched Multimegabit data service (SMDS) CoS
• Frame relay CIR
• IP ToS

14
Characterizing Services

• Service performance requirements
• Reliability and availability
• Capacity
• Delay

15
Characterizing Services

• Reliability
• A measure of the systems ability to provide
  deterministic and accurate delivery of
  information
• Completion time (e.g. 15 seconds for a sales
  transaction, round trip delay lt 40 ms for
  real-time banking transaction)
• Accuracy correct info received

16
Characterizing Services

• Capacity
• A measure of systems ability to transfer
  information
• Bandwidth the theoretical capacity of one or
  more elements in the system (amount of resources
  available)
• e.g. SONET OC-3c circuit 155.52 Mbps excludes
  protocol overhead or T1 (DS1) 24x64kbps1.536
  Mbps (1.544 Mbps if including overhead)
• Throughput the realizable capacity of the system
  or its components (how much user data can be
  transmitted in a stated period of time)
• e.g. SONET OC-3c circuit 80-128 Mbps (measured)
• The bandwidth of an Ethernet is 10Mbps, the
  throughput for a particular user depends on how
  busy the network is.

17
Capacity

• Capacity
• Often measure in bits or bytes per secondc
  bytes per second (1 byte8 bits)
• Packets (frames, cells) per second when data are
  transmitted in packetC in packets per second
  (PPS) (compare to customs/s in queuing theory)
• Packet length may be fixed or variable

18
Capacity and service rate

• Consider the arrival process as Poisson
• Approximation compressed voice/image, web
  traffic(?) with data rate r in bits per
  second?r/(packet length in bits)
• Variable exponential distributed ? M/M/1 with
  service rate µR
• Fixed M/D/1
• ? lt µ

Possibly combined sources Offered traffic ?
µ C (PPS)
r (bps)
c (bps)
Buffer large enough (or no significant queuing)
19
Capacity and service rate

• Example a capacity line with c56 kbps,average
  packet size128 bytesC56x103/(8x128) 55 PPS

M/M/1
C (PPS)
c (bps)
20
Capacity and service rate

• Example a capacity line with c56 kbps,fixed
  packet size48 bytesC56x103/(8x48) 146 PPS

M/D/1
R (PPS)
r (bps)
21
Characterizing Services

• Delay a measure of the time differences in
  transmission of information across the system
  (source to destination).
• Propagation delay
• Transmission delay
• Queuing delay
• Processing delay
• Latency Delay incorporating application
  processing and task completion times

22
Network Analysis Process

• Requirements Analysis
• User requirements
• Application requirements
• Host requirements
• Network requirements
• Flow Analysis
• Flows
• Data sources and sinks
• Flow models
• Flow boundaries
• Flow distributions
• The flow specification

23
(No Transcript)
24
Requirement Analysis Concepts

• Understanding the design environment
• Identifying, gathering and understanding system
  requirements and their characteristics,
• Developing performance thresholds
• Determining specified services for the network
• Helps the designer better understand the probable
  behavior of the network being designed

25
Generic System
26
User requirements

• The system should adapt to the user environment
• Provide quick and reliable information access and
  transfer
• Offer quality service

27
User requirements

• General requirements
• Timeliness
• Interactivity
• Reliability
• Quality
• Adaptability
• Security
• Affordability

28
User requirements

• We also need to know
• User numbers
• User locations
• Expected growth

29
Application requirements

• Application is the second component on the
  Generic System (see previous diagram)
• Services in the network can be described by the
  requirements
• Reliability
• Capacity and
• Delay

30
Application requirements
31
Application Types

• Mission-critical application
• Deterministic and/or guaranteed reliability
• Controlled-rate application
• Specified capacity
• Real-time (and possibly interactive) application
• Specified delay

32
Reliability

• A measure of the systems ability to provide
  deterministic and accurate delivery of information

33
Reliability

• Loss of reliability may lead to
• Loss of revenue/customers
• Unrecoverable information or situation
• Loss of sensitive data
• Loss of life
• Best effort delivery/predictable or bounded
  reliability

34
Capacity

• Controlled-rate applications
• Voice
• Non-buffered video
• Some teleservice applications
• Best effort applications
• Traditional file transfer (e.g. ftp over TCP)
• Capacity impacts delay

35
Delay

• Sources
• Propagation
• Transmission
• Queuing
• Processing, routing, etc.

36
Transmission/Queuing Delay
M/M/1
µ C (PPS)
?

• M/M/1 Delay1/(µ-?)
• Average packet size in bits S
• ?r/S, µc/S
• Delay1/(c/S- r/S)
• Hence delayS/(c-r)

c (bps)
r (bps)
37
Transmission/Queuing Delay
M/D/1
µ C (PPS)
?
c (bps)
r (bps)

• M/D/1 Delay1/2(µ-?)
• Fixed packet size in bits S
• ?r/S, µc/S
• Delay1/2(c/S- r/S)
• Hence delayS/2(c-r)

38
Transmission/Queuing Delay
µ C (PPS)
?
c (bps)
r (bps)

• µgtgt?, delayS/c ?transmission delay, no queuing
  delay
• Data chunk as a big packet data burst S in
  bitsdelayS/c

39
Transmission Delay

• Depends on transmission speed (bandwidth or
  throughput)

40
Queuing Delay
µ C (PPS)
?
c (bps)
r (bps) average rate or busy hour average rate

• Queuing delay depends on the traffic patterns
• Traffic source modeling detailed traffic pattern
• Traffic pattern (coarse)
• Peak rate
• Average rate - random, but predictable average
• Bulky transfer or busy hour average rate

41
Source Model Parameters

• Burstiness how frequently a source sends traffic
  (a value between bigger than or equal to 1)
• Source activity factory from 0 to 1, can be
  expressed in percentage
• Utilization average fraction of transmission
  capacity used by a source
• Recall Delay1/(µ-?)S/(c-r), µc/S, S packet
  size, c capacity

42
Delay-specific applications

• Real-time applications
• Strict timing relationship between source and
  destination
• Real-time speech, non-buffered video
• Non-real-time applications
• Interactive applications
• BatchAsynchronous applications

43
Delay-specific applications

• Non-real-time applications still time limits
  apply, however, destination will wait
• Interactive applications
• telnet
• ftp
• Web applications
• Asynchronous applications
• E-mail

44
Application Delay Types
45
Typical Application Traffic Attributes
46
Application Groups

• By developing application groups, often we can
  more quickly and easily determine general
  performance characteristics by mapping the
  application to one of the groups

47
Application Groups

• Command and control/telemetry applications
• High-performance delay and reliability,
• mission-critical and/or real-time
• Example aircraft control
• Visualisation applications
• High performance capacity or delay
• Real-time and/or controlled rate
• Example weather modelling, aeronautics, medicine

48
Application Groups

• Distributed computing applications
• High performance delay,
• Possibly interactive
• Exampleparallel computing, cluster, distributed
• Applications for Web access, development and use
• Delay sensitive but not high performance
• Bulk data transport
• Not high performance
• Example ftp

49
Application Groups

• Tele-service applications
• High performance capacity, delay, and/or
  reliability
• Requires multicast backbone (mbone)
• Examples Teleconferencing, telemedicine,
  teleseminars
• Operations, administration and maintenance
• High reliability
• DNS, SMTP, News (NNTP), address resolution,
  network monitoring and management, network
  security, and systems accounting

50
Host requirements

• Types of hosts and equipment
• Generic computing devices
• PCs, MACs, UNIX workstations etc.
• Servers
• Specialised equipment
• A bank of infrared sensors, a parallel system
  supporting a large database search engine,
  supercomputers, mainframes, parallel and
  distributed computing systems, data gathering and
  processing system

51
Host requirements

• Location information
• Locations of existing and expected
• Hosts
• Servers and
• Specialised equipment
• Helps to determine flow characteristics

52
Network requirements

• Existing networks and migration
• Constraints, size, relocating, network-layer and
  support services
• Functional requirements
• Management and security
• Financial requirements
• Level of funding available
• Enterprise requirements
• Phone, fax, voice and video, as well as data

53
Requirement Analysis Guidelines

• Requirement analysis process model
• Gather requirements
• Develop service metrics to measure performance
• Characterising behaviour
• Develop performance thresholds
• Distinguishing between service requirements

54
Requirement Analysis
55
Gathering and listing requirements

• Determining initial conditions
• Type of design new, modification, analysis etc.
• The scope of the design size, distance, number
  of sites
• Initial design goals
• Outside forces political, administrative,
  financial
• Working with users

56
Gathering and listing requirements

• Developing Service metrics to measure performance
• availability and reliability
• MTBF (mean time between failures) obtained under
  stress test
• MTTR (mean time to repair)
• MTBSO (mean time between service outage)
• Availability the amount of time the system is
  working when compared to the measured life time
  of the system
• High reliability three nines?Ai99.9
• Unavailability (probability of failure) Ui1-Ai
• Average number of failures in time tt/(MTBF
  MTTR)t/MTBF
• Recoverability or stability, in terms of

57

• availability and reliability
• Availability the amount of time the system is
  working when compared to the measured life time
  of the system
• Example MTBF1000 hoursthe number failures in
  one (8760 hours)8.76 failures or 0.024 failures
  per day ?2.4 failures per day for a 100 nodes
  throughout the network!
• Serial network As(A1) (A2) UsU1A2A1U2U1U2
• Parallel network ApA1A2-A1A2
  UpU1U21-A1-A2A1A2

A1
A2
Serial
A1
Parallel
A2
58

• availability and reliability
• Reliability distribution of time between
  failures (exponential distributed with meanMTBF
  - Markovian failures)
• Prob. (tltt)1 - e-t/MTBF
• Reliability prob. That there is no failure prior
  to te-t/MTBF
• Example MTBF10,000 hours, t1year8760 hours
  Reliabilityexp(-8760/10000)41.64

59
Gathering and listing requirements

• Service metrics for capacity
• Data rates Peak data Rate (PDR), average data
  rate, minimum data rate
• Busy hour traffic
• rBH of bits received in the busy hour/3600
  (bps)
• Example rBH 5 PPS, S128 bytes, c9600 bps
• rBH 5x128x8x3600/3600 5120 bps
• UtilizationBH5120/960053.3
• Data size burst size and duration
• Design for peaks busy hour basis or busy
  minutes basis or burst basis

60
Gathering and listing requirements

• Service metrics for delay
• End-to-end, round-trip or system delay
• Latency
• Delay variation
• Timeliness

61
End-to-end delay

• Circuit switching
• Setup delaytransmission delaypropagation delay
• Example 2Mbits, c128kbps, transmission
  time2/12815.6 s
• Message switching (assume no queuing delay)
• sum of transmission delay prop.delay
• Example 2Mbis 2/1282/1.54416.9 s
• Packet switching
• total message delay prop.delay
• total message delay(S/c)h(M-S)/c (?M/c)S
  packet size of the first packet M message size
  c capacity h number of hops ( of nodes-1)
• Example 2Mbits, packet size1024bits,
  c56kbpstot.msg del.(1.024/56)(2)(2-1.24)/56
  35.8 s

Transm. time
time
128kbps
1.544Mbps
Prop.time
message sw.
Transm. time
time
128kbps
128kbps
Prop.time
packet sw.
62
Bandwidth limited vs. latency limited

• Propagation delay ? distance, medium
• queuing plus transmission delay (M/c)/(1-?) ?
  average utilisation, M message (burst) size
• queuing plus transmission delay gt? bandwidth
  limited application -- the destination begins
  receiving data before the source has completed
  transmission of the message (burst)
• queuing plus transmission delay lt? latency
  limited application -- the source has finished
  the message (burst) before the destination begins
  receiving any data
• not too efficient
• c for queuing plus transmission delay ?
  bandwidth/latency crossover point
• the bigger the ?, the bigger the
  bandwidth/latency crossover point

63
Gathering and listing requirements

• Configuring and measuring service levels
• Simple Network Management Protocol (SNMP)
• Common Management Information Protocol (CMIP)
• Ping, traceroute, pathchar (round-trip delay,
  per-link capacity, path traces), NIST NET,
  net-snmp/mrtg (Multi-Router Traffic Grapher)

64
Characterising behaviour

• Usage patterns
• Frequency and duration of application sessions
• Number of simultaneous sessions
• Application behaviour
• Data sizes
• Frequency and time duration data passes the
  network
• Traffic flow characteristics
• Degree of multicasting

65
Developing Performance Measures

• Reliability requirements
• Availability
• Reliability (MTBF/MTBSO/MTTR)
• Error or loss rates
• Per link or circuit
• Between network equipment end-to-end
• Thresholds are set for reliability

66
Developing Performance Measures

• Delay requirements
• Interaction delay
• 10 30 seconds
• Human response time (HRT) time threshold when
  users begin to perceive delay
• 100 ms
• Below HRT Users do not notice delay
• Above HRT Users notice delay and as it grows the
  User becomes frustrated and loss of productivity
  may result
• Network propagation delay
• Depends on technology and distance (lt 1 sec)

67
Developing Performance Measures

• Delay
• System responsiveness (SR) HRT/TCT when HRT/RTT
  1
• System responsiveness HRT/(RTTTCT) when
  HRT/RTT lt 1
• HRT Human response time
• RTT Round-trip time
• TCT Task Completion time
• SR lt 3, a low degree of responsiveness
  (interactive bulk)
• SR gt 3, a high degree of responsiveness
  (interactive burst)

68
Developing Performance Measures

• Delay
• Burstiness (peak data rate)/(average data rate)
• End-to-end (system) delay
• Propagation, queuing, transmission, I/O,
  switching, and processing
• Delay variation (DV)
• 1 - 2 of end to end delay
• End-to-end delay 40 ms, DV 400-800 microsecond

69
Developing Performance Measures

• Capacity requirements
• Estimating data rates

70
Developing Performance Measures

• Supplemental performance requirements
• operational suitability can the design be
  configured, monitored and adjusted
• supportability can the design be kept performing
  as designed
• confidence is the network able to deliver data
  without error or loss as designed