1. Wireless Mobile Communication: From Circuits to Packets
Fouad A. Tobagi
Stanford University
European Wireless Conference
Barcelona, February 26, 2004

2. A Brief Historical Perspective

3. The Telephone Network Circuit Switching 64 Kbps circuits 1878 Toll
Trunk
Switch
Toll
Trunks
Customer
Central
Office
Circuit Switching
64 Kbps circuits
1878

4. The ARPANet Packet Switching Statistical multiplexing 1969 C SWITCHING
REMOTE
TERMINALS
LOCAL
ORPHAN
COMPUTER
FACILITY
SWITCHING
EQUIPMENT
HIGH-SPEED LINE
Packet Switching
Statistical multiplexing
1969

5. The ALOHA System 1970 Terminal Multi-access Channel (f1)
Broadcast Channel
(f2)
Central
Computer
Terminal
1970

6. Packet Radio Network Ground Packet Radio System (GPRS) 1973
SOURCE
DESTINATION
STATION
Ground Packet
Radio System
(GPRS)
1973

7. Public Data Networks X.25 Packet switching Virtual circuits
Approved by CCITT in 1976
1976

8. Local Area Networks 1980 IEEE 802.5 (Token Ring) 4 Mb/s 16 Mb/s
. . .
LAN Segment
Station
IEEE (Ethernet)
10 Mb/s
IEEE 802.5
(Token Ring)
4 Mb/s
16 Mb/s
Bridge
1980

9. Campus Network Mid 1980’s o WAN Router FDDI (100 Mb/s) Ring Segment
Bridge
16 Mb/s
Ring Segment
Subnetwork
10 Mb/s
WAN
Mid 1980’s

10. A Global Data Network
Campus
Network
WAN
Mid 1980’s

11. The Internet Protocol (IP)4 8 16 19 24 31
VERS
IHL
TYPE OF SERVICE
TOTAL LENGTH
IDENTIFICATION
FLAGS
FRAGMENT OFFSET
TIME TO LIVE
PROTOCOL
HEADER CHECKSUM
SOURCE IP ADDRESS
DESTINATION IP ADDRESS
OPTIONS
DATA
PADDING
20 bytes
1975
Datagram format

12. Data Network Applications
Resource sharing
Remote login
Electronic mail
News
File transfer

13. Wireless Voice Networks
PSTN
Cellular
Network
Wireless voice communication
Full mobility management solution
1990

14. A Growth Spurt Making the Internet Public Advent of the World Wide Web
Data traffic growth (50-300% per year)
Making the Internet Public
Advent of the World Wide Web
1995-present

15. Wireless Data Networks
Data Network (Internet)
Wireless
LANs
Wireless data communication
No mobility management
1997+

16. Toward a Converged Network

17. One Network for Each Type of Traffic
PSTN
Data Network (Internet)
1995-present

18. Toward a Converged Network
Forces at work:
Ubiquity of the internet (50M users in 4 years)
Deregulation of telecommunications industry
Market readiness for new communications services and applications
Advances in technology:
Semiconductors, photonics, wireless

19. Access Network Technologies
Residential Access Networks
Access
POPs
DWDM
CMTS
DSLAM
Base
Station
L2 Switch
fiber TP
cable modem
optical DWDM ring
xDSL
wireless
Ethernet

20. New Applications Collaboration Distance Learning News Home shopping
Karaoke
Medical diagnostics
Pay-per-View
Training
Investment
Video Conferencing
Corporate Communication
Banking
Factory Floor Reference
Telephony

21. Shift
PSTN
Converged Network
Data Network

22. Converged Network Packet-based IP-based
Statistical multiplexing efficiency for data traffic
Flexibility to meet varying requirements of new applications
Open client-server paradigm in management, control, and services
IP-based
Ubiquity of IP
Advances in associated protocols

23. New Applications

24. New Applications Communication among people News and entertainment
Education and training
Information retrieval
Commerce
Corporate communication
Health care
Advertising, publishing
Factory floor reference
...

25. Communication Among People
Voice communication (voip, IP telephony)
Ubiquity of the internet
Alternative to telcos
Integration with other applications
New functionality
Conferencing (made easier)
Storage (record, play-back, index, edit, integrate…)

26. Communication Among People
Video Conferencing
A picture is worth a thousand word
facial expressions, gestures, reactions…
Same advantages as with voice communication
Insertion of video clips
Fly-on-the-wall
Quality
Collaboration
shared white board
more frequent meetings

27. News and Entertainment
News in all its forms (paper, audio, video, web, combination; Live and stored)
Selectivity (on-line, by profile…)
Accessibility without frontiers
Urgent notification
Linkage among various sources
Archival material

28. News and Entertainment
Movies and TV programming
Movie-on-demand (pay-per-view)
Large selection
Full VCR functionality
Live broadcasts (sports, weddings, …)
Wider audience
Interactive games

29. Education and Training
Distance learning
Distance independence
Asynchronous learning
Time independence
Flexible curriculum
Flexible pace
Monitoring

30. Business Applications
Information kiosks
Corporate communication
Factory floor reference
Banking
Home shopping
E-commerce
Publishing
Etc...

31. Medical Applications
Medical imaging
Tele-surgery!
Health education

32. New Traffic Types Voice Video Images Stream oriented Delay sensitive
High bandwidth ( Mb/s)
Images
High data volume

33. Characteristics and Requirements
Bandwidth Requirement
Latency Requirement
Types of traffic
Traffic Pattern
ms. (interactive communications)
Voice Telephony
Stream-oriented symmetric
6-64 Kb/s
Video Video conferencing
Entertainment (Movie-on-demand)
VOD applications
Stream-oriented
 symmetric
asymmetric
ms.
minutes (near VoD)
seconds
1-2 Mb/s
20 Mb/s (HDTV)
4-6 Mb/s (MPEG2)
Data Web browsing
E-commerce
Other ( , file transfer)
Random & bursty
asymmetric
unpredictable
< 1 sec. (interactive, time sensitive)
No real-time requirement
10 mb/s (peak)
1 Mb/s (average)

34. New Networking Requirements
Bandwidth
Latency
Multicasting
Integrated services
Roaming
Nomadic access
Seamless handover
“Enable high performance data communications for mobile workforce, whether on company premises, in the field or at home” (Paul Henry)

35. Service-oriented Internet

36. Sources of Requirements
Users
Application developers
Providers
Network Functions and Architecture

37. Users Requirements High quality of service Mobility
Support effectively new types of traffic (voice, video)
Low latency
Good quality
Differentiated services
High network availability and reliability
Simplicity in using network
Low cost
Security and privacy
Mobility

38. Service Provider Requirements
Ease of network configuration and resource allocation
Customer care management
Usage tracking and accounting
Policy management
Flexible network solutions
To meet evolution and growth

39. Application Developer Requirements
Rapid development
Open architecture
Isolation from network details
Standard common service-oriented support functions
Ease of integration with other applications

40. A Three-level Logical Architecture
Major Applications
Telephony & Personal Communications
Web Access & E-commerce
Distance Learning
News &
Entertainment
Network Planning & Provisioning
Session Management
Customer Care Function
Content Mgmt
Security
Multicast- ing
QoS Delivery
Session Establish- ing
Customer Care
Networking Resource Management
Policy Management
Usage
Tracking & Account- ing
Directory Services
Networking Resource Directory
Multicast Group Directory
Authentication & Encryption Directory
Policy Directory
Customer Directory
Infrastructure
Hosts
Optical Network Elements
Layer 2 Switches
Layer 3 Routers
Gateways
Monitoring Devices

41. Wireless Mobile Data Communication

42. Two Independent Efforts
The internet world:
Mobile IP
The cellular voice network world:
General packet radio service (GPRS)

43. The IP World IP Addressing: Hierarchical Aggregate entries ScalableSW SW
MAC Addressing
flat address space
Individual address
Subnet

44. Mobile IPv4 CN R R R SW SW HA FA
Foreign
Network MN
Home Network

45. Problems With Mobile Ipv4
Triangular routing
Route optimization?
Deployment problem:
Availability of FA in foreign networks
Hampered by use of private ipv4 addresses and network address translators
Ingress filtering
Mechanisms for authentication and authorization are specific to mobile ipv4
Separate protocol for registrations (using UDP)

46. Mobile Ipv6
Mobility signaling and security features integrated as header extensions
Address auto-configuration:
Stateful using dhcpv6
Stateless (no need for FA), using router advertisement and router solicitation ICMP messages, and combining foreign network prefix with MH interface identifier
Built-in route optimization
Biding updates sent to HA and CN (biding requests and biding acknowledgements)

47. General Packet Radio Service
Link
Layer
Mobility
1. Attach
2. Activate
PDP
Context
Source: A. Samjani, “General Packet Radio Service [GPRS]”, IEEE Potentials,Volume: 21 , Issue: 2, April-May 2002 Pages:

48. Integration, Not Convergence

49. Wireless LANs and Cellular Data
“The Wireless LANs standardization and R&D activities worldwide, combined with the recent successful deployment of WLANs in numerous hotspots, justify the fact that WLAN technology will play a key role in the wireless data transmission”
Source:A. Salkintzis, C. Fors, R. Pazhyannur, Motorola, “WLAN-GPRS integration for next generation mobile data networks”, IEEE Wireless Communications,Volume: 9 , Issue: 5 , Oct. 2002, Pages:

50. Wireless LANs and Cellular
“A cellular data network can provide relatively low-speed data service over a large coverage area. On the other hand, WLAN provides high-speed data service over a geographically small area. An integrated network combines the strengths of each.”
Source:A. Salkintzis, C. Fors, R. Pazhyannur, Motorola, “WLAN-GPRS integration for next generation mobile data networks”, IEEE Wireless Communications,Volume: 9 , Issue: 5 , Oct. 2002, Pages:

51. Wireless LANs and Cellular
“Public WLANs can hardly be seen as competing with true mobile data systems. However, they can be deployed as a complementary service to GPRS/UMTS, owing essentially to their bandwidth/cost ratio.”
Source:Public Wireless LAN for Mobile Operators, WLANs beyond the entrerprise, Technology White paper by Alcatel

52. WLAN-GPRS Integration (Loose Coupling)
WLAN deployed as an access network complementary to the GPRS
Network. Uses only subscriber databases in GPRS.
Source: A. Salkintzis, C. Fors, R. Pazhyannur, Motorola, “WLAN-GPRS integration for next-generation mobile data networks”, IEEE Wireless Communications,Volume: 9 , Issue: 5 , Oct. 2002, Pages:

53. WLAN-GPRS Integration (Tight Coupling)
WLAN Connected to GPRS core network as a radio access network
Source: A. Salkintzis, C. Fors, R. Pazhyannur, Motorola, “WLAN-GPRS integration for next-generation mobile data networks”, IEEE Wireless Communications,Volume: 9 , Issue: 5 , Oct. 2002, Pages:

54. WLAN-GPRS Integration
“Typically, no user intervention would be required to perform the switchover from WLAN to GPRS. Moreover, the user would not perceive this handover. When the user moves back into the coverage of a WLAN system, the flow would be handed back to the WLAN network”
Source: A. Salkintzis, C. Fors, R. Pazhyannur, Motorola, “WLAN-GPRS integration for next-generation mobile data networks”, IEEE Wireless Communications,Volume: 9 , Issue: 5 , Oct. 2002, Pages:

55. Will Convergence Ever Happen?

56. Wireless Access Network
Internet Evolution
Fixed Wired
Infrastructure
End-to-End
Quality of Service
Wireless Access Network
Routing to mobile users

57. Is the Internet Ready for VoIP?

58. VoIP System and Impairments
Delay impairments
Interactivity
(<150ms)
Echo
Speech quality
Compression
Packet Loss
Network
talk
silence
Voice source
Sender
talkspurt
Encoder
Packetizer
Delay jitter
Loss impact depends on: loss duration, rate, PLC
Interactivity impact depends on: m2e delay, task
Echo: hybrid, acoustic. Echo cancellation, delay.
Quality requirements
The network may introduce impairments
packet loss, delay, delay jitter, echo
Some of them can be overcome at the end system, to some extent:
Speech quality: low loss rates, small loss durations (<60-90ms)
Interactivity needs end-to-end delay <150ms
Depacketizer &
Playout buffer
Decoder &
concealment
Receiver

59. VoIP Quality Measure Mean Opinion Score (MOS)
Speech Transmission Quality according to user satisfaction
4.3
4.0
3.6
3.1
2.6 1
5.0
Desirable
Acceptable
Best (very satisfied)
High (satisfied)
Medium (some users dissatisfied)
Low (many users dissatisfied)
Poor (nearly all dissatisfied)
What is the right way to assess VoIP quality?
Ideally we should have humans all day rating real calls. All paths, all hours… Infeasible
Our automated way. Emulates humans’ opinion and predict MOS. Define MOS
Other studies (many of which are by AT&T) have mapped: impairment conditions> MOS rating
[ITU-T G.109 recommendations states that a rating R in the ranges [90,100], [80,90], [70,80][60, 70], [50,60] corresponds to best, high, medium, low and poor quality
respectively [very satisfied, satisfied, some users dissatisfied, nearly all users unsatisfied, not recommended]. A rating below 50 indicates unacceptable quality.
[ITU-T G.107/annex b defines translations R-MOS]
P.800 defines MOS.
Not
recommended

60. Loss Impairment for G.711 10ms and 20ms
Here I am showing results from three different sources. A lot of work done by AT&T speech research.
The starting point Ie(loss=0) depends on the Codec.
The slope of the curve Ie(loss rate) depends on the parameters: codec, PLC, loss model (bursty/uniform, packet size=unit of loss)
Sources from references: [Emodel, AT&T, Voran, Gruber]. They all agree!
Bottomline is that above 10%, nothing can be done. Therefore the clusters we saw 10-80%, cannot be concealed.

61. Delay Impairments Interactivity impairment: Echo impairment:
Depends on “task” and total delay
Echo impairment:
Depends on echo cancellation and total delay
Notation:
EL=echo loss in dB
Task= type of conversation

62. Assessment of Backbone networks
Why backbone networks only? In practice the paths may have more segments.
Say the names of providers.
1.We have measurements collected over backbone networks
2. This is still interesting because:
Backbones are an important part of the end-to-end path (long distance calls, call going through both a data and a circuit switched network.)
It is revealing that we identified problems on these backbone networks, that are usually considered over-provisioned and not
to be a problem. The bottlenecks are considered in other parts of the network.
Factors in growth: (1) use of access in the population (2) moving from modems to DSL (3) new applications demanding higher rates (video)
It is unclear how the growth will continue. (2) also means that the traffic becomes burstier – needs more overprovisioning fro QoS.
Circuit in the core will help to have higher capacity network (instead of overprovisioning).
Even in a very over-provisioned network, occasional spikes are an issue. Control or routing traffic
Delay is inherently now low: packetization+algorithmic. Slow access (wireless)+ propagation coast to coast (23-45ms). They bring you up to 100ms.
Wireless
Access

63. Internet Backbone Measurements
Probe based measurements (RouteScience).
Backbone networks of 7 major ISPs.
AND
EWR
SJC
THR
ASH
Our study is limited to backbone networks. In practice the paths may have more segments. Why only backbones?
1. We have measurements collected over backbone networks
2. This is still interesting because:
Backbones are an important part of the end-to-end path (long distance calls, call going through both a data and a circuit switched network.)
It is very interesting that we identified problems on these backbone networks, that are usually considered over-provisioned and not to be a problem.
Well, this may be the case for TCP based traffic, but it is not for VoIP, as we will see.

64. Packet Loss Characteristics
Rare sporadic single packet loss
Repetitive single packet loss
“Clips”: consecutive packets lost
19-25 packets lost
Long clips (outages)
Duration: 10s of seconds - 2 minutes
Usually preceeding changes in the fixed part of the delay
Often happen simultaneously on more than one path of a provider

65. Example of Repetitive Single Loss
EWR-P3-SJC, Thu 7:20 (UTC)
4 paths of an ISP
48 hours period
1 packet lost every 5 sec on average (0.2% loss)
Explain the plots – we will see these kind of plots all the time
Y axis: delays of individual packets. X axis could be sequence number, but I put send time to show the duration. Delay 0 means the packet is lost.
The CCDF plots are plotted at log-y to easily compare with the exponential distribution that should have a straight line CCDF at logy plot.
Loss rate 0.2% is not much for voice. One packet lost every 4-5sec on average. For the entire measurement period.
total loss: 36096, Single packet loss: 33023, 4 clips ~200, One clip 1978 packets
It is not important for voice, but it may have an effect for TCP.

66. Example of a Clip
EWR-P2-SJC, Thu, 13:50
230 ms clip

67. Example of Outage Outage of 112 sec ASH-P7-SJC, Wed, 4:00
change in fixed delay
reverse path: outage 166sec
next day same time, both paths

68. Packet Loss Characteristics
Clustered packet loss
High loss rates (10-80%) for up to 30 sec
Synchronized with similar events on other paths
Precede or follow changes in delay

69. Example of clustered packet loss
EWR-P6-SJC, Wed 3:20 (UTC)
9.4% loss: 141 single packets in 15 sec
Accompanying increase in delay
Synchronized with events on 3 other paths of the same provider
10 minutes
zoomin (1)
zoomin (2)

70. More Complex Loss Events
EWR-P2-SJC Wed 06/27/01 3: EWR-P2-SJC Thu 06/28/01 20:10

71. Packet Delay Characteristics
Low delay variability
High delay variability
Mixed behavior

72. Example 1: Low Delay Variability
SJC-P7-ASH Wed 6/27/01
Minimum delay= 77 ms
High delay, high delay variability. Load varies during the day.
Loss not a problem on this path, except for 2 times: 1 sec and 800ms clip.

73. Example 2: High Delay Variability
THR-P1-ASH
Minimum delay= 77 ms
High delay, high delay variability. Load varies during the day.
Loss not a problem on this path, except for 2 times: 1 sec and 800ms clip.

74. Example 3: Mixed Behavior
SJC-P2-ASH Thursday 06/28/01
Minimum delay= 77 ms
High delay, high delay variability. Load varies during the day.
Loss not a problem on this path, except for 2 times: 1 sec and 800ms clip.

75. Delay Components Fixed delay: Delay variability:
Path connecting sites:
Fixed Delay
In the east coast only ms
From/to Colorado
28 – 78 ms
Coast-to-coast ms
Fixed delay:
Delay variability:
Mostly in the form of spikes
less frequently congestion
There are consistent patterns per provider/path/time
width
delay in ms
(send) time in sec
distance
fixed delay
peak
clustering
Spike patterns are expected because these are backbone paths (overprovisioned).
What makes it interesting is:
- the size of spikes that we are going to see.
- the fact that there are systematic patterns that we are going to see.
- confirmation there are not many slow varying components (important for the playout) –this was expected (overprovisioning)
Because of the triangular shape, the distribution of the peaks is the same with the distribution of all delays

76. Simple Spike From P7 (A)

77. High Spike From P1 (B)

78. Cluster of Spikes From P4 (C)

79. Non Triangular Spike From P5 (D)

80. Effect of Delay Jitter
A spike means that packets arrive bunched-up
Action to handle a spike:
Send
Receive
2. gap
3. loss
4. adjust rate
1. buffering
Play
delay in ms
(send) time in sec
These spikes are not a problem for data applications. But they are an issue for voice
Send
Receive
Play ?
Playout scheduling ?

81. For more information
A. Markopoulou, F. A. Tobagi and M. Karam, “Assessment of VoIP quality over Internet backbones,” Proceedings of the IEEE INFOCOM 2002, New York, June 2002.
F. A. Tobagi, A. P. Markopoulou and M. J. Karam, “Is the Internet Ready for VoIP?” Proceedings of the 2002 Tyrrhenian International Workshop on Digital Communications – IWDC 2002, Capri, Italy, September 2002.
A. P. Markopoulou, F. A. Tobagi and M. J. Karam, “Assessing the Quality of Voice Communication over Internet Backbones,” IEEE/ACM Transactions on Networking, Vol.11, No. 5, Ocotber 2003, pp

82. VoIP Over 802.11 Wireless LANs

83. VoIP Performance Capacity of a voice-only 802.11 network:
Maximum number of simultaneous voice calls that can be supported -
For a given MOS requirement
Distribution of voice quality across users taking into account channel conditions (frequency selective fading)
Not realistic channel – very simplified

84. 802.11 Key Features CSMA/CA No collision detection
“Listen before you talk”
No collision detection
Frames are positively acknowledged
Collisions and errors in transmissions
Retransmissions
Random delay
Packet may eventually be dropped
AP must backoff after every transmission
Stations can’t differentiate between collision & error

85. Network Scenario Single Basic Service Set (BSS) 802.11(b) at 11 Mb/s
N wired users
Single Basic Service Set (BSS)
802.11(b) at 11 Mb/s AP
N wireless users (‘stations’)

86. An Upper Bound on Capacity
Analysis assuming no collisions and no errors
Encoder
(data rate)
Voice Data Per Frame
10ms
30ms
50ms
G.711 (64kbps) 6 18 26
G.729 (8kbps) 7 22 35
mbps
Don’t get 8x capacity using 1/8 rate!
For maximum capacity, use G.729 with 50ms voice per packet

87. Where Does the Time Go? G.711 (64kbps), N = 18, 30ms speech/packet
Pure TDMA system (ignoring control overhead) using the same frame format would avoid idle time, => 33% increase in capacity.

88. How Tight Is the Upper Bound?
Simulation with no errors
Simulation (analysis)
Effect of collisions is very low…
Voice Data per frame
10ms
30ms
50ms
G.711
6 (6)
17 (18)
25 (26)
G.729
7 (7)
21 (22)
34 (35)
Effect of collisions is low, but:
Number of users is of the order of Cwmin. (31).
… for this scenario!

89. Observations AP Access Point is a bottleneck Very few collisions occur
Frames dropped in AP downlink queue
Very few collisions occur
Typically, probability of collision for any given transmission ~ 3% at AP
Failure is sudden
quality at (Nmax + 1) is very poor AP
Capacity is not function of MOS
No capture effect as seen in other protocols
Not surprising because… (collisions, access point has half the traffic, but is unlikely to collide with itself; exp. B/o is conservative)

90. How many collisions does a frame incur?
G.711
G.729
Voice Data per frame (ms) 10 30 50
Capacity 6 17 25 7 21 34 AP
1.6
2.8
3.9
2.7
3.5
3.7
Stations
2.0
5.3
8.7
3.2
6.1
8.9
Observe, big difference between AP, station
Doesn’t affect capacity that much – double counting, doesn’t cost full transmission
Probability of transmission colliding (%)

91. Retransmissions How many packets incur x collisions?
Collision resolution is very effective when using exponential backoff..
e.g. at AP, p(2 colls| 1 coll) = 4.1%, p(3 colls | 2 colls) = 7.1%
At non-AP p(2 | 1 ) = 10.6% p(3|2) = 7.9%
Max retransmissions = 7; clearly we could reduce this without any impact

92. Capacity with Delay Constraints
Target MOS; e.g., 3.6, 4.0
Playout deadline:
150ms causes no degradation in MOS [source: ITU E-model]
Maximum acceptable loss rate
e.g. for G.711, 10ms packets, MOS 3.6, maximum acceptable loss rate is 4.9% [source: ETSI TR v.2.1.1, 2002]
Delay budget for wireless network + packetization
Take distribution of delay for all packets

93. Delay CCDF – G.711
Explain CCDF : probability that any frame incurs delay > d
Steep downward curves are best
Recall delay does not include wired delay
*This is the aggregate statistics for all wired-to-wireless traffic*
Example – 1% loss, 55ms delay budget
pick packet size according to delay!

94. Tradeoffs & Limitations
Packet Size:
Larger packets increase capacity, but have high packetization delay cost
Harder to conceal loss of longer packets
G.729 vs. G.711:
G.729 requires 5ms look-ahead at encoder
G.711 has lower capacity with no delay constraints
G.729 has lower intrinsic quality (3.65 vs 4.15 for G.711)

95. Delay-constrained Capacity*
Assume optimum packet size selection
g.729 cannot achieve 4.0 mos
Very similar results
Not senstive to MOS (v. sharply falling curves, loss goes from 0 -> high v. quickly)
Throughput bound vs. delay bound e.g. 50ms frames useless if your budget is 40ms
* In error-free channel, with optimal packet size selection

96. Observations Capacity highly sensitive to delay budget
May be worth increasing delay budget, sacrificing MOS for higher capacity
Wireless Network Delay is low for N < capacity
Very similar results for G.729/G.711
Low sensitivity to MOS requirement
Optimal packet size can be obtained considering packetization delay only

97. Capacity with Delay Constraints
What happens if we consider frame errors?

98. Channel Errors - Intuition
Channel errors decrease capacity and increase delay:
More retransmissions require more time on the medium
Each packet requires (on average) more transmissions

99. Channel Errors – Approach
Constant BER model
All stations + AP experience equal channel conditions
BER from 10-6 to 2 x 10-4
Capacity for BER  10-3 is 0
PHY header assumed to be received correctly
Transmitted at 1Mbps
All MAC frame errors are detected, but cannot be corrected
Like, everyone equidistant from the AP (station to station channel not really important, as long as they’re not hidden)
Constant BER – no coding – really, there is independent symbol errors (8 bits/symbol)
Slightly more complex than just requiring re-tx, but not going into that now.

100. Capacity for MOS = 3.6
G.729
G.711

101. For more information
David P. Hole and F. A. Tobagi, “Capacity of an IEEE b Wireless LAN Supporting VoIP,” Proceedings of the International Conference on Communications, ICC 2004, Paris, France, June 2004. 

102. VoIP Over IEEE a

103. System Parameters ETSI indoor channel A. Packet size – 154 bytes.
Typical office environment with non-line of sight NLOS.
RMS delay spread - 50 ns.
Maximum delay spread ns.
Packet size – 154 bytes.

104. Average VoIP Quality Ignoring Fading

105. Average VoIP Quality With Fading

106. Call Quality Distribution (No Retransmissions)

107. Call Quality Distribution (up to 3 Re-tx)

108. Packet Error Rate Distribution

109. For more information
Olufunmilola Awoniyi and F. A. Tobagi, “Effect of Fading on the Performance of VoIP in IEEE a WLANs,” Proceedings of the International Conference on Communications, ICC 2004, Paris, France, June 2004.

110. Role of Layer 2 Technologies in Mobility Management

111. The IP World IP Addressing: DNS Hierarchical Aggregate entries
Scalable
DNS R R R SW SW
MAC Addressing
flat address space
Individual address
Subnet

112. Tracking and Routing in the Internet
Directory
(DNS)
Gives a fixed IP address
Persistent and Complete
Layer 3
Route to the user subnet
- Static
- Scalability by address aggregation
Layer 2
Learns about user
Highly Dynamic
Effect of collisions is low, but:
Number of users is of the order of Cwmin. (31).

113. Mobile IP CN R R R SW SW HA FA
Foreign
Network MN
Home Network

114. Wide-area Mobility Via Mobile IPHA CH SW R R SW GW R FA SW R GW
..Triangle routing, frequent IP address changeover, slow handoffs

115. Proposed IP Wireless WorldGW SW
Overlay R
Extend one subnet to large areas (many hundreds of square km)

116. MobiLANe Concept
Extend one subnet to large areas (many hundreds of square km)
Dynamic DNS G SW
Internet
Gateway
Cells
Switch
Mobile IP Tunneling
MobiLANe

117. Issues What structure should the network have?
What are appropriate protocols for user tracking and routing?
What is the optimal size of the network?
What is its reliability?

118. Tracking and Routing Issues
Passive learning and flooding
Fast mobility => learning quickly obsolete => more flooding => not scalable to many users (bandwidth overload at switches and possibly links)
Tracking along spanning trees => slow updates for movement between certain sections of the tree
Flat address space
Large, unstructured databases at nodes in the spanning tree (especially close to the root). Address distribution via multiple spanning trees helps, but only by a constant factor at best
Inherent tradeoff of memory technology (fast access => small size; Large size => slow access)

119. MobiLANe Exp learn Multicast
Instead, use explicit learning (GARP like protocol)
Combine learning with selective multicast to reduce database size and improve worst-case updates
Concept similar to m-regional matching (Awerbuch-Peleg 1995) but made practical
Use cache hierarchies to optimize lookups
Exp learn
Multicast

120. For more information
C. Hristea and F. A. Tobagi, “IP Routing and Mobility,” Proceedings of IWDC 2001, Taormina, Italy, September 2001, Springer Verlag LNCS, Vol
C. Hristea and F. A. Tobagi, “A network Infrastructure for IP Mobility Support in Metropolitan Areas,” Computer Networks, Vol. 38, pp , February 2002.
C. Hristea and F. A. Tobagi, “Optimizing Mobility Support in Large Switched LANs,” Proceedings of the IEEE International Conference on Communications, ICC 2003, Anchorage, Alaska, May 2003.

121. Ad Hoc Networks

122. A Scenario
sensor
Pocket PC
PDA
Mobile
phone
Handheld
Computing
capability
GPS& location
GPS & location
Computing capability
Source: A. Helmy, USC, “Service Provisioning in Large-scale Infrastructure-less Wireless Networks”

123. Ad Hoc Networks
Made its debut for applications in military tactical operations (Packet Radio Network, Survivable Radio Network, etc.)
Made possible by the use of packet switching
Easy deployable
Wider area coverage without the need for infrastructure
Many applications scenarios

124. IEEE

125. Source: IEEE 802.20 Requirements Document – Ver. 9, November 5, 2003

126. Source: IEEE 802.20 Requirements Document – Ver. 9, November 5, 2003

127. The Air-Interface (AI) shall be optimized for high-speed IP-based data services operating on a distinct data-optimized RF channel.
The AI all shall support interoperability between an IP Core Network and IP enabled mobile terminals and applications shall conform to open standards and protocols.
The MBWA will support VoIP services. QoS will provide latency, jitter, and packet loss required to enable the use of industry standard Codec’s.
The systems must be designed to provide ubiquitous mobile broadband wireless access in a cellular architecture.
allowance for indoor penetration in a dense urban, urban, suburban and rural environment.
Source: IEEE Requirements Document – Ver. 9, November 5, 2003