Download Teletraffic tools for performance analysis

Document related concepts

Computer simulation wikipedia , lookup

Network science wikipedia , lookup

Transcript
Teletraffic tools for performance analysis of
telecommunication systems
Trial Lecture
Anne Nevin
Centre for Quantifiable Quality of Service in Communication Systems
(Q2S)
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Teletraffic tools for performance
analysis of telecommunication
systems
Presentation outline:
• Introduction to Network performance
• Network Performance measures
• Teletraffic tools for performance analysis
• Conclusion
1
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Presentation outline:
• Introduction to Network performance
• Network Performance measures
• Teletraffic tools for performance analysis
• Conclusion
1
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
A telecommunication system is a
system consisting of the necessary
equipment for seamless transfer of
information
user
electromagnetic transmission of information
A
(wire pair, coaxial cable, fiber optics and radio)
user
B
2
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
A telecommunication system is a
system consisting of the necessary
equipment for seamless transfer of
information
user
A
user
underlying network
B
2
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Performance is the ability of a system
to provide the resources needed to
deliver its service
user
A
user
Network: underlying
Collection of
systems that
network
exchange data among themselves
B
3
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Performance is the ability of a system
to provide the resources needed to
deliver its service
user
A
Information services Electronic trading
Network: Collection of systems that
Communication
exchange
data amongservices
themselves
Real-time communication
user
B
Juridic Exchange
3
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
The users do not care what the
underlying network is. They only care
about the quality of the service
Real-time communication
4
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
The users do not care what the
underlying network is. They only care
about the quality of the service
Real-time communication
4
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Performance will be tightly linked to
the network’s capabilities of:
user
A
Traffic handling
Security
user
Dependability
B
5
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Performance will be tightly linked to
the network’s capabilities of
user
A
Traffic handling
user
B
5
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Traffic handling
limited network
capacity
user demand
Links, routers,,
switches etc…
Performance
6
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Performance and network capacity
Performance
Network Capacity
6
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Performance and network capacity
Performance
Requirement
Network Capacity
6
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Performance and network capacity
Performance
Requirement
Over provisioning and poor
utilization of resources
Network Capacity
6
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
The use of network capacity should be
optimized
Net
Traffic
(stochastic
user demand)
Net
Network
Performance
Net
Network
capacity
7
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
To say something about
network performance we need
well-defined performance
measures
8
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Presentation outline:
• Introduction to Network performance
• Network Performance measures
• Teletraffic tools for performance analysis
• Conclusion
9
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Two fundamental information transfer
modes
Packet switched
networks
(wired and wireless)
Circuit switched
networks
Next generation
Internet ?
(mobile and fixed)B
1876:
A. G. Bell
1969 Internet 1991 www
Arpanet
2010
time
10
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Circuit Switched network (e.g PSTN): a circuit lasts for the
duration of the call
11
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Circuit Switched network (e.g PSTN): a circuit lasts for the
duration of the call
11
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Circuit Switched network (e.g PSTN): a circuit lasts for the
Call set-up and release delay
duration of the call
Mobile:
handover
Probability of blocking
11
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Circuit Switched network (e.g PSTN): a circuit lasts for the
Call set-up and release delay
duration of the call
Mobile:
handover
Probability of blocking
Packet switched network (Internet), traffic is segmented into
packets. No dedicated circuit and capacity is shared.
11
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Circuit Switched network (e.g PSTN): a circuit lasts for the
Call set-up and release delay
duration of the call
delay
Mobile:
handover
Probability of blocking
Packet switched network (Internet), traffic is segmented into
packets. No dedicated circuit and capacity is shared.
Queue of packets
11
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Circuit Switched network (e.g PSTN): a circuit lasts for the
Call set-up and reales delay
duration of the call
Mobile:
handover
Probability of blocking
Packet switched network (Internet), traffic is segmented into
packets. No dedicated circuit and capacity is shared.
Throughput
Packet loss
Queue of packets
Packet delay
www.q2s.ntnu.no
Packet loss pattern
Packet delay variation
11
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Performance measures related to
traffic handling:
Circuit switched network
(call-level)
Packet switched network
(packet-level)
• Call set up and release delay
• Probability of end-to-end
blocking
• Probability of unsuccessful
handover(mobile)
• Packet delay and delay
variation
• packet loss rate and packet
loss pattern
• Throughput
means, variances, or their
probability distributions
12
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Presentation outline:
• Introduction to Network performance
• Network Performance measures
• Teletraffic tools for performance analysis
• Conclusion
13
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Teletraffic tools are used to analyse
performance related to traffic in a
telecommunication system
14
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Three categories of Teletraffic tools
• Measurement tools
• Simulation tools
• Mathematical tools
15
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Three categories of Teletraffic tools
• Measurement tools
• Simulation tools
• Mathematical tools
15
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Measurement tools are used to carry
out measurements on real systems
or prototypes
Software or
hardware
16
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Measurement tools can be grouped
into two types: Passive measurement
tools and active measurement tools
Software or
hardware
17
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Passive measurement tools
Software or
hardware
Software or
hardware
18
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Passive measurement tools
Example: DAG cards (Data Acquisition and
Generation), Both off-line and on-line capabilities.
Privacy issues
Software or
hardware
Software or
hardware
18
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Drive testing: Passive measurement tool for
wireless and mobile network performance
measurements
Performance measures:
coverage, capacity, hand-over, call-setup, blocking
probability in mobile networks
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Active measurement tools
Artificial
probe traffic
Software or
hardware
Characterize
probe traffic
Software or
hardware
Should not
exceed 1%
19
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Active measurement tools
Artificial
probe traffic
Software or
hardware
Characterize
probe traffic
Software or
hardware
Example: PING (measure round-trip delay) and
PATHCAR tool (estimating link capacity and delays)
Telephone system: dummy subscribers
19
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Measurement tools are the only means
of verifying that performance is
according to the specifications
• What is the quality of the measurements?
• There may be large stochastic uncertainties and these
must be considered
• Measurements may be collected from large-scale
systems (e.g. Internet). Are we even measuring what
we think we are measuring?
• What is the quality of the analysis of the data?
• Measurement analysis can become time consuming and
a complete sensitivity analysis is almost impossible
20
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Three categories of Teletraffic tools
• Measurement tools
• Simulation tools
• Mathematical tools
21
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
A simulation tool is used to analyse
the performance of a simulation
model
Traffic
User demands Data generated
from real traffic
Artifical data
Structure
Strategy
Hardware
Software
Topology,
capacity
Wireless, wired
Transfer mode
Figrue ”Teletraffic Engineering and Network Planning” by Villy B Iversen
www.q2s.ntnu.no
Priorities?
Scheduling
Admission Control
22
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Simulation tools can be classified
according to their level of details
• General purpose programming (C, C++, Java, Python
etc).
• Simulation programming languages which are
specifically designed for simulation purpose (e.g Simula)
• Simulation libraries: Provide building blocks needed to
make a simulator. (e.g Java-sim, DEMOS)
• Specialized tools: Build-in components that can be reconfigured. (e.g the Network Simulator, NS).
23
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Simulation tools make performance
analysis flexible
• A good simulation model can be quite detailed and
realistic and provides control on all underlying
parameters.
• The significance of the individual parameters may be
difficult to grasp
• A simulation model is not general. Every scenario must
be simulated.
• Both stocastic uncertainies and quality of data analysis
must be considered.
• The simulation model must be verified using a real
network
24
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Three categories of Teletraffic tools
• Measurement tools
• Simulation tools
• Mathematical Analysis tools
24
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Mathematical performance analysis is
carried out using a mathematical
representation of the system
MMRP
r
ON
b
a
OFF
Topology,
capacity
Wireless, wired
Transfer mode
Poisson
Parameter
Traffic
exampels:
Characteristics
Mean
are modeled
Variance
Distribution
Structure
Strategy
Hardware
Software
Priorities?
Scheduling
Admission Control
a mathematical model
www.q2s.ntnu.no
25
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
When doing a mathematical analysis,
the mathematical models are the
tools
Analytical solution:
Non-analytical solution:
A closed form solution
gives an exact functional
relation between model
parameters and
performance measures
Does not have a closed
form and must rely on
numerical methods. The
result is then a long set of
numbers, not an equation
An ideal mathematical model mimics the real system
26
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
The Erlang-B formula
Erlang 1st formula
Net
Offered
Traffic
Net
Blocking
Probability
www.q2s.ntnu.no
Number
of
Net
servers
(links)
27
19
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
The Erlang-B formula
Erlang 1st formula
Poisson arrival
Net
Offered
Traffic
Call duration distribution:
X, mean: E(X)
Net
Blocking
Probability
Number
of
Net
servers
(links)
19
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
The Erlang-B formula
Erlang 1st formula
Poisson arrival
Net
Offered
A=λE(X) Traffic
Call duration distribution:
X, mean: E(X)
Net
Blocking
Probability
Number
of
Net
servers
(links)
19
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
The Erlang-B formula
Erlang 1st formula
Poisson arrival
Net
Offered
A=λE(X) Traffic
Call duration distribution:
X, mean: E(X)
Net
Blocking
Probability
Number
of
Net
servers
(links)
n
19
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
The Erlang formula (circuit switched)
27
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Queuing Models
Kendall’s notation (both loss and waiting time systems)
A/B/n/k
A=Arrival process
B= service process
n= number of servers
k= number in system
A=B=M=Markov process
Arrival
process
Number of queuing places
28
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Queuing Models
Kendall’s notation (both loss and waiting time systems)
A/B/n/k
A=Arrival process
B= service process
n= number of servers
k= total number in system
A=M=Markov process
Erlang 1st formula
= Probability of call blocking
M/M/n/n
Poisson arrival
μ
28
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Queuing Models
Kendall’s notation (both loss and waiting time systems)
A/B/n/k
A=Arrival process
B= service process
n= number of servers
k= number in system
A=B=M=Markov process
Erlang 2nd formula
= Probability of waiting
M/M/n/
28
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Queuing Models are also used
in packet network
Kendall’s notation (both loss and waiting time systems)
A/B/n/k
A=Arrival process
B= service process
n= number of servers
k= number in system
29
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Queuing Models are also used
in packet network
Kendall’s notation (both loss and waiting time systems)
A/B/n/k
A=Arrival process
B= service process
n= number of servers
k= number in system
A=M, B= G= general or M=D=Markov
Queue of packets
29
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Queuing Models are also used
in packet network
Kendall’s notation (both loss and waiting time systems)
A/B/n/k
A=Arrival process
B= service process
n= number of servers
k= number in system
A=M, B= G= general or M=D=Markov
M/G/1
Queue of packets
29
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Call arrivals follow a Poisson process. Packet
arrivals do not
Some challenges
• Applications will have different rate characteristics
• Correlation and long-range dependence issues
• Taking advantage of the statistical multiplexing gain
30
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Mathematical models used in packet networks
Equivalent Bandwidth: For the single node analysis.
Captures the statistical multiplexing and determines the
minimum required bandwidth to meet a certain
delay/loss probability
Network calculus: ‘A system theory’ for the internet. The
min- plus algebra transforms the system into a linear
system which is analytically tractable.
It is then possible to carry out the analysis of delays and
backlogs in the network as if the network is a black box
31
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Mathematical models give an exact
functional relationship between
performance measures
• If the system changes the model may no longer hold
• There is a trade-off between simplicity and accuracy of
the model (closeness to the real system)
• Given two models with the same accuracy, the simplest
one to solve is the most desirable
• A model that is accurate but very complex can be
difficult to understand and analyze and can also pose
computational problems
32
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Presentation outline:
• Introduction to Network performance
• Network Performance measures
• Teletraffic tools for performance analysis
• Conclusion
1
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
Conclusion
• Teletraffic tools are used to quantify performance related
to traffic in a telecommunication system
• The tools can be grouped into three categories
Measurement
tools
Simulation
tools
Flexible, understand
behavior, validation
Validate and
estimate parameters
Mathematical
analysis
tools
Exact relations
accuracy issues
33
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems
” It is difficult to see the impact of a collection of
mathematical results as you try to master them; it is only
after one gains the understanding and appreciation for
their application to real-world problems that one can say
with confidence that he understands the use of a set of
tools”
[Leonard Kleinrock, in ”Queuing Systems” 1975]
www.q2s.ntnu.no
Anne Nevin, Teletraffic tools for performance analysis of telecommunication systems