COSACS 99

1
MULTI-MEDIA TDMASATELLITE NETWORKS
Claude Bélisle Patrick Larouche Peter
Andreadis http//www.crc.ca

• COSACS - 99
• 28 Octobre 1999

2
Presentation Outline

• Background
• Simulator Design
• Traffic Modeling
• Traffic Management
• Simulation Examples

3
Satellite Communications Market

• Todays satellite communications sector
  represents 2.3 of the total telecommunications
  market.
• This number is expected to increase to 6 over
  the next ten years with over 46 millions
  subscribers in 2008 and revenue of about 50
  billions.
• This increase will result primarily from the
  emerging market for high-mobility remote access
  broadband service and internet. Satcom will be
  complementing and competing the terrestrial
  wireless and fiber optics networks.
• Canada is a world leader in telecommunications
  and the national policy on Connecting Canadian
  will strengthen this role.

4
Network Architecture Concept
5
Applications

• Voice
• Fix and mobile services
• Data Service
• Messaging, Asset tracking, e-mail, File
  transfer, Imagery
• Video
• Tele-conference, Television
• Real-time and non real-time
• Internet access
• Web browsing,
• Multi-media
• Tele-learning, tele-medecine

6
Satcom Benefits

• Ubiquitous coverage
• Bandwidth on demand
• Support for mobile communications
• Rapid deployment of terrestrial infrastructure

7
Proposed Multi-Media Satcom

• Anik F2 Canada GEO Ku/Ka
• Skybridge USA LEO Ku
• Astrolink USA GEO Ka
• Cyberstar USA GEO Ka
• Teledesic USA LEO Ka
• N-Star Japan GEO Ka
• Euroskyway Europe GEO Ka
• SES Astra Europe GEO Ku/Ka

8
Networks over Satellites- The Issues /1-

• Channel error rate
• Retransmission of packets (TCP protocol),
• TCP congestion control protocol,
• Propagation delays
• Resource access requests,
• Sliding window, acknowledgements.
• Bandwidth
• Expensive resource which needs careful management.

9
Networks over Satellites- The Issues /2-

• On-board processing,
• Circuit switch versus Packet switching.
• Multiple satcom and terrestrial protocols

• Adaptation of terrestrial network protocols
• Efficient resource management techniques
• Multi-service terminals

10
Project Objective

• Design and implement a software tool for
  
• modeling, simulating and analyzing
• advanced satellite communications networks

11
Project Scope

• Three main activities are being investigated
• Analysis of satellite load under given scenarios,
• Efficient Medium Access and Congestion Control
  techniques,
• Terrestrial network protocol performance over
  satellite links.

12
Sponsoring Organizations

• This work is being sponsored by
• The Canadian Space Agency under the Advanced
  Satellite Communications program,
• The Department of National Defence through the
  Research and Development organization (CRAD).

13
Presentation Outline

• Background
• Simulator Design
• Traffic Modeling
• Traffic Management
• Simulation Examples

14
Network Architecture Simulation Concept
15
Satcom Network Architecture
16
User Traffic Functionality

• Generates the communication traffics
  representative of the different applications
  (voice, video, data, internet, multi-media)
• Encapsulate the traffic in the appropriate
  terrestrial protocols
• TCP,
• IP
• ATM or Ethernet
• Sends the traffic to the ground terminal

17
Ground Terminal Functionalities

• Protocol adaptation ( terrestrial ? satellite ),
• Simple encapsulation (extra overhead)
• Modification of the existing overhead
• Terminal characteristics
• transmission rates,
• number,
• locations
• Resource Manager
• Request bandwidth for itself to the satellite
  resource controller
• Act as satellite resource controller.

18
Payload Functionalities

• Satellite characteristics
• Transponders data rate,
• Air interface protocol,
• On-board processing
• multiplexing / demultiplexing,
• coding / decoding,
• packet switching / circuit switching,
• beam switching, intersatellite links.
• Resource Manager

19
Presentation Outline

• Background
• Simulator Design
• Traffic Modeling
• Traffic Management
• Simulation Examples

20
Traffic Models

• Traffic models are essential in order to design
  and predict the performance of the network.
• Access protocols may be quite different for
  different traffic types.
• Optimization techniques, for bandwidth usage,
  heavily depends on traffic characteristics.
• Traditional traffic models fail to explain the
  behavior of the existing networks in the presence
  of todays traffic.
• Todays traffic is more bursty and exhibits
  greater variability
• e.g. MPEG encoded video.

21
Traffic Types

• Voice
• Data (e-mail, imagery, file transfer...)
• Video (real-time and non-real time)
• Internet Web browsing
• Multi-media

22
Traffic Model Challenges

• Represent real traffic
• Flexible to model traffic variability
• Parsimonious
• Computationally efficient

23
Traffic Models

• A lot of research is being done to model
  communication traffics.
• In Canada, main research groups EMS
  Technologies, Carleton U. Concordia U. Ottawa U.
  Waterloo U.
• The most popular models are
• Poisson process
• Markov Modulated Poisson Process (MMPP)
• On - Off WWW
• Self - similar

24
Poisson Process

• Good description for human initiated process
  (telephone calls)
• Not appropriate for multi-media applications
  since it lacks temporal distribution and can not
  capture the traffic burstiness.
• Mainly used as a standard reference model.

25
Markov Modulated Poisson Process (MMPP)

• Introduces dependence in the random sequence.
• Can potentially capture traffic burstiness.
• Is a doubly stochastic Poisson process
• Packet interarrival rate (?) and sojourn times
  (?) are exponentially distributed.
• Mainly used for voice and video applications
• With ?2 0, 2-MMPP is called On-Off process.

?1
?2
?1
2 - MMPP
?2
26
On - Off WWW

• Based on Internet WWW real traffic, monitored at
  GTE lab.
• WWW document transmissions are not entirely
  initiated by the user. Therefore, a Poisson
  distribution is not valid.
• Model can be described by the distribution of
  three random variables
• Inter-arrival time of requests during ON period
  (Weibull distribution)
• Duration of the ON period ( Weibull distribution)
• Duration of the OFF period (Pareto distribution)
• The packet length follows a Pareto distribution.

27
Self-Similar Marginal Transform Process SS-MTP

• Some classes of traffic still do not follow the
  previous models (coded video - JPEG and MPEG).
• The traffic is bursty but shows both short and
  long-range dependence as it may appears the same
  (i.e. similar to itself) after a certain period.

28
Traffic Model Implementation
29
Presentation Outline

• Background
• Simulator Design
• Traffic Modeling
• Traffic Management
• Simulation Examples

30
Traffic Management

• Regulate the traffic in order to
• obtain high bandwidth utilization,
• avoid network congestion,
• provide an acceptable quality of service to the
  users,
• and all that at low service cost.
• Must therefore consider
• traffic models (flow, rates, scalability),
• bandwidth allocation (U/L, D/L and ISL),
• congestion control.

31
Traffic Manager Design
Traffic Multiplexing
Admission Control
Feedback
Congestion Control
Bandwidth Allocation
32
Traffic Multiplexing

• The variable nature of the communications traffic
  enables multiplexing of a number of sources in
  order to allocate the network bandwidth more
  efficiently.
• Traffic from various sources may be multiplexed
  based on
• Data rates, QoS, destination.
• Aggregate data are sent on the network.

Voice
Data
Video
Web
M-M
33
Traffic Multiplexing

• Very high source aggregation, such as Internet
  backbone,
• The traffic load will tend to smooth out over
  large periods.
• The best MAC strategy may be a dedicated channel,
  if delay is critical and cost not a factor.
• Medium source aggregation (10 - 100 Mb/s), such
  as in the case of large companies, or an internet
  service provider (ISP).
• Traffic load will remain bursty.
• Dynamic bandwidth allocation would be preferred
• Small source aggregation ( lt 10 Mbps), such as
  personal communications
• Dynamic bandwidth allocation is essential

34
Admission Control

• This function deals essentially with the routing
  of the information based on destination (earth
  coverage, spot beams, intersatellite links)
• Monitoring of the traffic on each beam is
  essential to prevent overflow in one of the link.
• The Admission control deals with the initial
  request of the terminal. Minimum and peak date
  rates are negotiated and a minimum QoS is
  guaranteed.

35
Admission Control

• Admission control can be based on various
  techniques
• Fixed Assignment
• Capacity is locked and no negotiation needed.
• Simplest form but maybe costly and bandwidth
  inefficient
• Random Access
• There is no negotiation and terminals transmit at
  random on specific time slots pre-allocated.
• Inefficient for medium and large number of users
• Demand assignment
• Capacity is given based on some request

36
Bandwidth Allocation

• Due too the bursty nature of the traffic
  re-negotiation may become necessary within a
  connection.
• Three techniques can be used
• No knowledge
• When traffic increases, a request is sent to the
  resource manager requesting more capacity.
• If the process is delay sensitive, this may not
  be efficient especially if the RM is on the
  ground.

37
Bandwidth Allocation

• Feedback based
• When bandwidth requirements changes, a request is
  sent to the resource manager for bandwidth
  modification. Response is obtained and
  modifications occur.
• Decentralization of the task could be achieved if
  information regarding the load of the network was
  fed back to the ground terminal. Based on the
  load, the terminal could automatically reserve
  extra capacity and inform the resource manager
  later.

38
Bandwidth Allocation

• Prediction based
• The ground terminal could look at the
  characteristics of the traffic stream and
  anticipate future bandwidth requirements.
  Request could then be made prior to the actual
  requirement time.

39
Congestion Control

• Used to prevent overflow of the network load
  capacity.
• Rate control,
• Packet discarding,
• Priority traffic multiplexing

40
Traffic Management IssuesWho is the boss ?

• Satellite
• Quickest response time but most expensive at less
  flexible to protocol changes and special user
  groups.
• Ground terminal
• More flexible but slower due to longer delays.
• Distributed
• More complex but may be most effective depending
  on traffic load.

41
Simulation Example

• Assume
• 105 voice terminals
• Each terminal requires 2 of the bandwidth per
  connection.
• Each user is connected 30 of the time
• Network management protocol Connection
  reservation

42
(No Transcript)
43
(No Transcript)
44
(No Transcript)
45
(No Transcript)
46
ANY QUESTIONS ?
Web site http//www.crc.ca e-mail
claude.belisle_at_crc.ca
47
(No Transcript)
48
(No Transcript)