Bandwidths is changed Frame size conforms to

2010 Bi-Weekly Cores Lab Meeting
Impact of Highly Active Primary Users on IEEE 802.22
Network: A Single Cell Case
Jihoon Park, Pål Grønsund, Przemysław Pawełczak
2010 Bi-Weekly Cores Lab Meeting
Motivation
 IEEE 802.22 is a Wireless Regional Access Network standard developed by IEEE
since early 2006
 Standard is still in the draft phase (latest version is 2.0, July 2009)
 This is one of three currently available standards that focus on TV white spaces
 The other two are IEEE 1900.4 (no networking, just information sharing) and ECMA392 MAC and TV operation in the White Spaces
 IEEE 802.11af group has submitted its Project Authorization Request for new
standard: IEEE 802.11 in the TV white spaces
There seem to be no papers that would analyze this network in detail
 Whenever IEEE 802.22 name appears the system analyzed has actually nothing to
do with the real standard
 Example of papers: Zhao et al. (Tridentcom’09), Liu et al. (EMC’07), Hu et al. (Comm
Mag’07), Song et al. (Chinacom’08),
2010 Bi-Weekly Cores Lab Meeting
Question
Given
1. Realistic activity of Primary Users (channel bandwidths, signal levels, activity
patterns)
2. In the area with densely populated Primary Users
3. Detailed implementation of IEEE 802.22 (traffic types, admission strategies, frame
structure, modulation and coding types, bandwidth, subcarrier allocation and
channel sizes and channel numbers)
What is the average throughput and
delay of IEEE 802.22 network user
experience?
2010 Bi-Weekly Cores Lab Meeting
Approach
Investigation is composed of two parts
1.Analysis of steady state system behavior (throughput only) for a
simplified network (more in system model) for tractability reasons
2.Further investigation (throughput and delay) via extensive NS-2
simulations with presumably first in the world implementation of IEEE
802.22 stack
NS-2 simulations will be also used to see how severe the
simplifications of analytical model were and how well the analysis
follow NS-2 traces
2010 Bi-Weekly Cores Lab Meeting
Illustration
2010 Bi-Weekly Cores Lab Meeting
Preliminaries
 IEEE 802.22 has tons of similarities with existing IEEE 802.16e
IEEE 802.22
IEEE 802.16e
OFDMA channel profile
(MHz)
Air interface
6,7,8
Burst allocation
Linear
20,28,17.5,14,10,8.75,7,3.5,
1.25
OFDMA, OFDM, Single
Carrier
Two dimensional
Subcarrier permutation
Distributed (with enhanced
interleaver)
No
Adjacent/distributed
10 ms, superframe: 16
frames
No superframe, multipme
frame sizes: 2, 5, 10, 20 ms
MIMO
Frame size
OFDMA
STC, beamforming
2010 Bi-Weekly Cores Lab Meeting
Preliminaries: superframe structure
160 ms
...
Superframe n-1
Superframe n
Superframe n+1
10 ms
frame 0
Superframe
Preamble
Frame
Preamble
SCH
frame 1
Frame
Preamble
. . .
frame 15
Frame
Preamble
...
Time
2010 Bi-Weekly Cores Lab Meeting
Preliminaries: frame structure
...
frame n-1
frame n
DS subframe
US subframe
TTG
DS PHY PDU
Frame
FCH DS burst 1 DS burst 2
Preamble
DSMAP
USMAP
U
D
C
...
Ranging
slots
BW
request
slots
UCS
US PHY PDU
Notification
(CPE m)
slots
DS burst x
MAC PDU 1
MAC PDU 1
... MAC PDU y
CRC
US PHY PDU
(CPE p)
... MAC PDU k
Pad
MAC Header
MAC Payload
...
US burst
D Optional
C broadcast
D MAC PDU
MAC Header
...
frame n+1
Time
MAC Payload
CRC
Pad
Selfcoexistence
window
R
T
G
2010 Bi-Weekly Cores Lab Meeting
Preliminaries: frame structure
...
frame n-1
Time slot 0
frame n
Time slot
frame n+1
Adaptive
Time slot N-1
N time slots
Downstream
Subframe
Upstream
Subframe
...
Time
2010 Bi-Weekly Cores Lab Meeting
Preliminaries: frame structure
...
frame n-1
frame n
...
frame n+1
Time
10 ms
US-MAP
Ranging/BW request/UCS notification
Burst
Burst
Burst
60 subchannels
time buffer
Burst m
Bursts
Burst 2
Bursts
time buffer
DS-MAP
Burst 3
US-MAP
Burst n
DS sub-frame
US sub-frame
(smallest US burst portion on a given subchannel= 7 symbols)
RTG
Burst
TTG
Frame Preamble
more than 7 OFDMA symbols
(4 or 5 symbols when scheduled)
Burst 2
Self-coexistence window
Burst 1
Burst 1
UCD
DCD
FCH
26 to 42 symbols corresponding to bandwidths from 6 MHz to 8 MHz and cyclic prefixes from 1/4 to 1/32
2010 Bi-Weekly Cores Lab Meeting
Analytical Model: Assumptions
One base station only (no spectrum sharing among multiple IEEE 802.22 base
stations)
A bandwidth B of one TV channel is fully available to IEEE 802.22 network,
provided that no Primary User is actively transmitting
Bandwidth is divided into multiple (logical) sub-channels
Two types of Primary Users are considered
 Wireless Microphones (high variation in channel occupancy), occupies Z (currently
Z=1) channel; it can appear on any sub channel
 Other auxiliary device (low variation in channel occupancy) – tries to resemble a
TV transmission, occupies Y (currently Y=2) channels; it can also appear on any
sub-channel
Activity of two types of PUS follow a Poisson process
 Parameterized by individual arrival and departure rate
Transmission is slotted (a parameter of our model – one slot is one frame size)
 Spectrum sensing process is assumed to be non-perfect
 False alarm probability affects the throughput of IEEE 802.22
 Note that mis-detection does not, as given in the standard (detection is done per frame
basis)
2010 Bi-Weekly Cores Lab Meeting
Analytical Model: Assumptions
IEEE 802.22 users generate two types of traffic
 Elastic traffic (Variable Bit Rate - VBR), occupies X (X is a real number) logical
channels
 Non-Elastic Traffic (Constant Bit Rate - CBR), occupies Y (at the moment Y=1)
logical channels
 Both streams are Poisson, described by individual values of arrival and departure
Admission control strategy
 Whenever a VBR call occupies a channel and CBR call arrives, VBR must free
space for VBR call by “squeezing” the number of occupied channels to allow CBR
to access
 When any of PU occupies a channel both CBR and VBR must vacate its
corresponding sub-channels
 CBR: switches to idle channel, if nothing available then connections is being
buffered; no requirement on continuous channel availability, Y channels can appear
anywhere in the bandwidth B
 VBR: tries to squeeze the connection, if no space available then it is being
buffered; just like in CBR no requirement on continuous channel availability
2010 Bi-Weekly Cores Lab Meeting
Analytical Model: Limitations
 No adaptive modulation features considered (yet)
 No two-stage spectrum sensing considered (yet)
 No co-channel interference (should we?)
 Infinite number of users
 No other connection strategies considered (in relation to our previous
work), like just buffering or switching only
 Obviously only one cell considered
 The opposite requires designing of channel sharing strategies among
many IEEE 802.22 base stations
 Example: what to do when in one location only two full TV channels are
present with three base stations?
2010 Bi-Weekly Cores Lab Meeting
Analysis
 (nt , nw , nc , nv )
#PU class 1, #PU class 2, #CBR flows, #VBR flows
0  Ux  Ux,max ,0  Ux  kx  Ux,max ,Ux  kx  Ux,max Ux
P(Ut ,U w ,U c ,U v  Ut  kt ,U w  kw ,U c  kc ,U v  kv ) 
Pt (Ut  Ut  kt )Pw (U w  U w  kw | Ut )
Psu (U c  U c  kc ,U v  U v  kv | Ut ,U m )
Transition probability for PU class 1
Ut
Pt (U t  U t  kt )   ft (n | U t , t ,Ts )gt (kt  n | t ,Ts ), 0  kt  U x,max  U x
n0
Pt (U t  U t  kt ) 
Ut

ft (n | U t , t ,Ts )gt (n | kt || t ,Ts ), U t  kt  0
n |kt |
Ut

n0
mn
Pt (U t  U t  kt )   ft (n | U t , t ,Ts )  gt (kt  m | t ,Ts ),U t ,max  U t
2010 Bi-Weekly Cores Lab Meeting
Analysis, cont.
Recursive solution is needed to compute departure probability
Ts
ft (n | Ut , t ,Ts )   P(t1 | Ut , t ,Ts ) ft (n  1 | Ut  1, t ,Ts  t1 )
0
P(t1 | Ut t )  Ut t e
Ut t t1
, P(0 | Ut  n, t ,Ts  t1  t 2  K  t n )  e
(Ut  n) t (Ts t1 t 2 K t n )
2010 Bi-Weekly Cores Lab Meeting
Analysis results and model verification
•PU: TV and WM
•SU: CBR Only
•t = 5 /s, t = 3 /s
•w = w = 3, 30, 300 /s
•c = c = 100 /s
•Pf=0.1
•M = 4
•Batch
–size 10000
–num 100
–conf 0.9
0.65
0.6
System throughput (Mbps)
0.55
0.5
0.45
0.4
0.35
0.3
0.25
0.2
0
10
1
2
10
10
 (/s)
w
3
10
2010 Bi-Weekly Cores Lab Meeting
Analysis results and model verification
•PU: TV and WM
•SU: CBR Only
•t = 5 /s, t = 3 /s
•w = w = 3, 30, 300 /s
•c = c = 100 /s
•Pf=0.9
•M = 4
•Batch
–size 10000
–num 100
–conf 0.9
0.1
0.09
System throughput (Mbps)
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0
10
1
2
10
10
 (/s)
w
3
10
2010 Bi-Weekly Cores Lab Meeting
Analysis results and model verification
0.9
•PU: TV and WM
•SU: CBR and VBR
•t = 5 /s, t = 3 /s
•w = w = 3, 10, 30 /s
•c = c = 10 /s
•M = 4 (#channels)
•Pf=0.9
•Batch
–size 10000
–num 100
–conf 0.9
0.85
System throughput (Mbps)
0.8
0.75
0.7
0.65
0.6
0.55
0.5
0
10
1
2
10
10
 (/s)
w
3
10
2010 Bi-Weekly Cores Lab Meeting
Analysis results and model verification
•PU: TV and WM
•SU: CBR and VBR
•t = 5 /s, t = 3 /s
•w = w = 3, 10, 30 /s
•c = c = 10 /s
•M = 4 (#channels)
•Pf=0.1
•Batch
–size 10000
–num 100
–conf 0.9
0.1
0.09
System throughput (Mbps)
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0
10
1
2
10
10
 (/s)
w
3
10
2010 Bi-Weekly Cores Lab Meeting
NS-2 Simulations
 Adaptation of existing WiMax forum IEEE 802.16e NS-2 stack
 More than 20.000 lines of code
 What is currently implemented
 Spectrum sensing (single stage)
 Bandwidths is changed
 Frame size conforms to IEEE 802.22
2010 Bi-Weekly Cores Lab Meeting
Two stage spectrum sensing
2010 Bi-Weekly Cores Lab Meeting
Two stage spectrum sensing
t
t+1ms
t+5ms
DL
Subframe
UL
Subframe
DL
Subframe
UL
Subframe
Fast Sensing
Freq
Time
RTG = receive transmit gap
TTG = transmit transceive gap
This implementations uses some symbols at the end of the UL frame.
802.22 uses quiet periods starting from the end of the frame
2010 Bi-Weekly Cores Lab Meeting
Two stage spectrum sensing
Freq
Time
OFDMA
Frame
Fine Sensing
25 ms (5 OFDMA frames)
PS: fine sensing can be set to other values, but it must be a multiply of OFDMA frame length
2010 Bi-Weekly Cores Lab Meeting
NS-2 Model
2010 Bi-Weekly Cores Lab Meeting
NS-2 Model
CBR = packetSize_ * pps = packetSize_ * 1/interval_
CBR_802.22= 100 * 1/100 = 10 K
Constant
= 10 * 8 = 80 Kbps
Exponential on/off process:
CBR_WM= 100 * 1/100 = 10 K burst (on) = 2 sec
= 10 * 8 = 80 Kbps
idle (off) = 3, 4, 5, 6 sec
Exponential
on/off
process:
CBR_TV= 1000 * 1/10 = 10 K
burst
(on)
=
60
sec
= 10 * 8 = 80 Kbps
idle (off) = 60, 120, 180 sec
2010 Bi-Weekly Cores Lab Meeting
NS-2 Model
 CBR traffic, Wireless microphone only, On time: 2 s, off is a variable
Throughput
(Initial Simulations)
1,800,000
Throughput (bps)
1,600,000
1,400,000
1,200,000
1,000,000
800,000
600,000
400,000
200,000
0
1
2
3
Idle periods (seconds)
4
5