2 Related Work 7
4.2 Frame Format Definition
In this section, for the coexistence of cognitive radio network, we base the characteristic that includes spectrum sensing, sensing report, resource request, and channel assignment to define three frame formats: Association Request, Resource Request, and CR Parameter Set.
4.2.1 Association Request
Fig7. Frame Format – Association Request
For identifying SUs and WIFI users, we revise the reserved field in the capability information of 802.11 association request and denote CR-Pollable. CR-Pollable of 1 indicates the user is a SU, and CR-Pollable of 0 indicates the user is a WIFI user.
4.2.2 Resource Request
For SU’s sensing report and resource request, we define the new frame format and denote Resource Request.
14
Fig8. Frame Format – Resource Request
# of Channel Request
# of Channel Request indicates the resource request of the user. If user needs to report sensing result, it also guarantees its resource request can be exactly received by the AP.
# of Sensing report
# of Sensing report indicates the total channels that are sensed by the user. Channel count of 0 indicates that is a channel request without sensing result.
Sensing Result
Sensing Result indicates which data channel is sensed by the user and the channel condition of this data channel.
Sensing Timestamp
Sensing Timestamp indicates the time that the SU sense the data channel 4.2.3 CR Parameter Set
For spectrum sensing and channel assignment coordination, we define the new frame format and denote CR Parameter Set.
15
Fig9. Frame Format – CR Parameter Set Group count
Group count indicates the SU’s group id in the polling listing Group period
Group period indicates the number of SU’s group Channel Assignment
Channel Assignment indicates the channel assignment for SUs transmission.
Polling Listing
Polling Listing indicates that the number of SUs and which SUs would be polled in this SUPERFRAME.
RP (Resource Period) duration
Resource Period duration indicates the time that SU can send associated request and resource request to the AP.
16
4.3 PCF Access Delay
Notation:
: residual DCF duration
M: the number of WIFI users
r: saturation user transmission probability
: the average back-off time
When AP wants to broadcast the beacon for initialing the PCF mode, it maybe cause access delay because of that users use CSMA/CA scheme in DCF mode. That, it means the channel would be occupied by user at the moment.
Fig10. PCF Access Delay problem
It would cause the PCF mode delay, and also delay the time for AP to boardcast contorl message that include spectrume sensing listing and channel assignment.
Moreover, causing the sensing report delay would impact spectrum database update and then cause the spectrum evluation result. Therefore, we want to analysis the access delay and eliminate it. We assume every WIFI user are saturation(always have data to transmit), and we consider the condition at the moment of initial PCF mode:
Case1 : AP doesn’t have data to transmit : 1 – r
d1 = +( ) + ( ( ) ) ( )
17
The moment is in the previous backoff period, and AP waits till the backoff period finish. Then, If no user transmits data , AP waits for PIFS, and then start to broadcast beacon. If someone transmits data, the AP wait for data transmission time and PIFS, and then also start to broadcast beacon.
Case2 : AP has data to transmit : r
= ( ) ( )+ ( ( ) ) (
)
The time is not in the previous backoff period, and AP needs to transmit the data.
AP might successfully transmit data or cause collision. Then, the AP wait for PIFS and then broadcast beacon.
Then, the PCF Access Delay can be wirtten as below:
D = (1 – r) d1 + r d2 accommodate a data frame with a predefined probability , i.e,
( ) Successful : AP send the CTS message to the user Failed : AP ignore the RTS message
18
4.4 SU Coexistence problem
Notation:
: the slot time , : the propagation delay
: the probability that at least one user transmits his data
: the probability that only one WIFI user transmits his data
: the probability that only one SU transmits his data
: the number of WIFI users in the WIFI network
K: the number of SUs in the WIFI network
r: user transmission probability
: the size of associated packet
Ts: the transmit time of data packet
Tas: the transmit time of associated packet
Tc: the collision time caused by users data transmission at the same time
H: PHY header + MAC header
We firstly introduce the user authenticated/associated procedure in the WIFI network. The user state may be one of following three states:
(1) unauthenticated/unassociated (2) authenticated/unassociated (3) authenticated/associated
Fig11. The service state in the WIFI network
In state (1), user should use the DCF mode to send the authentication request to AP for acquiring the necessary security information till AP successfully receives the authentication request and send the authentication response to the user. After user successfully receives the authentication response from AP, he can go to the state (2).
19
Then, in state (2), user should use the DCF mode to send the association request to AP for verifying user capability till AP successfully receives the association request and send the association response to the user. After user successfully receives the associated response from AP, he can go to the state (3). In state (3), user can send data frame in the DCF mode. calculate the relationship between arrival rate and service rate of SUs. Finally, we can calculate the maximum number of SUs that can coexistence in the heterogeneous network, and it is a stable network.
Fig12.SU association Model
We assume that M WIFI users are saturation (always have data to transmit) and K SUs are all the state3 users and would back to DCF to send association request before timeout. Then, with the number of SU, WIFI user in the DCF mode, and user transmission probability, we can calculate the service rate for SUs that success association request to maintain user state. The data access scheme of WIFI user would impact the service rate of SUs, so we firstly derivate channel access time for Basic
20
RTS/CTS Access
= DIFS++RTS+SIFS++CTS+SIFS++H+E[P]+SIFS++ACK
= DIFS++RTS+SIFS++CTS+SIFS++ +SIFS++ACK
= DIFS++RTS+EIFS The arrival rate of SUs:
The service rate of SUs:
= ( )
= ( )
= ( )
( )
= ( )
= ( )
( )
= (
) ( )
Then, the relationship between arrival rate and service rate in a stable network is:
When the arrival rate is equal to service rate, we can find the maximum number of SUs coexistence in the network.
21
4.5 PCF duration
Notation:
T: the duration of SUPERFRAME
: the duration of PCF mode
k : the number of polling secondary users
: the one channel sample sensing time for SU
: the accuracy threshold of spectrum evaluation algorithm
22
: the number of available TVB channels
m : the number of require sensing samples per data channel
s : the probability of report error
: the number of slots in the RP duration
R(x , K) : The expect acquired request when we divide the time duration into x slots and K users adopt the random selection scheme
In our CR MAC architecture, we use the PCF mode to implementation the MAC architecture and we would discuss the duration in this section. Then, it is composed of four parts: Beacon, Collection Period, Resource Period, and CF_END. The duration of PCF mode can be written as below:
= + + + 1) Beacon and CF_END
Fig13. Beacon Frame Format
= + +
Fig14. CF_END frame Format
= + + 2) Collection Period
23
Fig15. The procedure in Collection Period
+ +
+ +
AP polls k SUs in this collection period, and . If AP successfully poll one SU, the poll time would include the poll time, SIFS time, and sensing report time. If report error occurs, AP waits for PIFS and continues to poll the next SU. Then, the CP duration can be written as below:
= k ( ) ( ) ( ) 3) Resource Period
Fig16. The procedure of Resource Period
= + +
For increasing the channel utilization, we have to guarantee the number of acquire resource requests exceed the number of available channels. The duration of RP can be written as below:
= + ∙ + 2 ∙
= ( ) 4) Quiet Period
According to the number of sensing sample for achieving the sensing accuracy,
24
one channel sensing time, and the number of sensing users, we could write the duration of QP as below:
= ⌈ ⌉ 5) Data Period
Then, the DP duration is according to the beacon transmission time and QP duration, and the duration of DP is written as below:
( )
Future, we could calculate the CR network throughput through the QP duration and the acquire request from the RP.
25
4.6 Random Selection Scheme
In this section, we would try to analysis the expect acquire requests when users randomly select one slot to send their request.
Notation:
G(a,b): the number of combinations when a users collide in b slots
X(m)={ , ,..., }, represents collsion user number in collsion slots When the collsion happens on one slot, it means that at least two users send their requests in the slot. Then: users, we firstly choose i users from K users and choose i slots that these users send his request from R slots. It also means that the residual K-i users happen collsion in some slots and the maximum number of collsion slots is ⌊( ) ⁄ ⌋ that the user number in every collsion slots is just two users. Therefore, we can wirte the equeation as below:
26
The expect acquire request is:
E[Q] = ∑
Finally, according to the realtion between R and K, the maximum acquire requests and maximum collision slots are different, we write the equation as below:
R ≥ K ,
E[Q] = ∑ ( ) ∑ ( )
⌊( ) ⁄ ⌋
( )
Because user number is less than slot number, the maximum acquire requests is K and the maximum collsion slots is ⌊( ) ⁄ ⌋.
Because user number is bigger than slot number, the maximum acquire requests is R-1. The maximum collsion slots is ⌊( ) ⁄ ⌋ because we should deduct the i occupied slots of acqire requests.
R < K , R ⌊ ⌋ ,
E[Q] = ∑ ( ) ∑ ( ) ( )
Because user number is bigger than slot number, the maximum acquire requests is R-1. The maximum collsion slots is because the maximum collsion slots would exceed the total slot number.
27
4.7 System Flow Chart
Fig17. System flow chart of proposed CR MAC
Fig.18. The association procedrue of SUs
28
Fig19. Specturm sensing and sensing report procedure
Fig20. Resource request issue and data transmission procedure
29
5. Throughput Analysis
We firstly base on the proposed CR MAC architecture to analysis the CR network throughput. Futher, we extend the throughput analysis equation of WIFI network include the impaction of SUs. That, we derivate the throughput equation from the throughput analysis of “ Performance Analysis of IEEE 802.11 Distribution Coordination Function”. The author proposes throughput analysis for saturation WIFI user throughput. Finally, we base on the throughput influence threshold of WIFI network and the accuracy threshold of spectrum sensing evalution algorithm to calculate the maximum system throughput that includes WIFI network and CR
R: the number of acquiring resource request in this SUPERFRAME
: the number of frame per data period
: the data rate of user j on the channel i : the number of sensing samples of channel i
( ) : the accuracy bases on the sensing time and sensing report number
30
, j = 1,2,…,K
∑ ≤ , Then, CR Throughput can be written as below:
( ) ∑ ( ) ∑
M: the number of WIFI users in the WIFI network K: the total number of SUs in the network
k: the number of SUs in the WIFI network
r: the transmit probability of user in the WIFI network Ts: the transmit time of data packet
Tas: the transmit time of associated packet
Tc: the collision time caused by users data transmission at the same time E[P]: the expected user data packet size
: the slot time
: the probability that at least one user transmits his data : the probability that only one WIFI user transmits his data : the probability that only one SU transmits his data
: the system throughput of WIFI network with k SU in the network : the system throughput of WIFI network
In the following, we would introduction how to calculate the WIFI throughput.
We assume there exist some WIFI users and both of them always have date to transmit (saturation) in the network. Then, there also exist some CR users that would send association request in the network. We derivate the throughput equation from the throughput analysis of [13] that the author proposes a throughput analysis equation for saturation WIFI users throughput.
Then, the WIFI throughput can be written as below:
= ( )
31
= ( )
= ( )
( )
= ( )
= ( )
( )
= (
) ( )
The WIFI throughput is according to the probability of that there are k SUs in the WIFI network
= ∑ (
) (
)
32
5.3
System Throughput
Notation
M: the number of WIFI users K: the number of SUs
N: the number of data channels
: the throughput influence threshold for WIFI network : the accuracy threshold of spectrum evaluation algorithm α: the sensing time to acquire one sample.
g: the number of channel and SU group : the number of SUs in the group
: the number of sensing channels in the group : the number of available TVB channels
m: the number of sensing sample per data channel s : the probability of report error
: the duration of DCF mode
: the duration of PCF mode
: the maximum duration of PCF mode
: the residual duration of PCF mode
: the maximum number of polling SUs in PCF mode
: the minimum number of polling SUs in PCF mode
: the maximum coexistence number of polling SUs in PCF mode W( ,M): the throughput of WIFI network
( ): the throughput of cognitive radio network z: the throughput of all the networks
33
Problem description:
We try to find the maximum throughput of CR network and WIFI network, and there is a tradeoff between CR throughput and WIFI throughput because of the PCF mode duration. The other tradeoff is between the channel sensing time and sensing report time. If more users sense the channels simultaneously, it would decrease the QP duration but increase the CP duration. Because the sensing time per sample (10ms) is greater than report time per user (500us), we have to decrease the QP duration as possible to acquire maximum throughput.
Assume:
We assume that we can query cloud for the number of sensing sample per channel m, the number of sensing channel , and the number of SU in every group. Without loss the generality, we also assume the number of sensing channel is less than the number of user in every group and the maximum number of polling SUs in PCF mode would less than the number of SUs per group:
Calculate Procedure:
We first calculate the maximum duration of PCF mode and the system throughput that is based on the influence threshold .Then, decrease the maximum duration of PCF mode, calculate the new system throughput, and check if increase the throughput or not till the duration is equal to zero. Finally, we can acquire the maximum system throughput and the duration of every period that is based on the sensing accuracy threshold and throughput influence threshold .
34
Fig21. flow chart of calculating procedure Input: , , α, N, K, M, , T
1) query cloud for the sensing samples per channel m, the number of sensing channels , and the SU groups by
2) the WIFI throughput
(M, K ) = ∑ (
) (
)
3) calculate the maximum duration of PCF,
, and T =
4) calculate the maximum polling number, and
=
⌊ ⌋ If ≥m, go to (5)
Else if m, = or decrease , go to (5)
35 Else, find the minimum number of polling users
⌈
36
6. Analysis Result
In this section, we would show the analysis result. We divide into three parts (1) SU coexistence (2) WIFI throughput with SUs (3) System throughput. The duration of SUPERFRAME is 400ms. The required sensing sample per channel is 5 to achieve the accuracy 0.9. The transmission rate is 11Mbps, basic rate is 1Mbps, and the average packet size of WIFI user is 1500 bytes. The TVB channel bandwidth is 6MHz, the channel SNR is 10dB. The packet size of BEACON is 464bits, CF_END is 160 bits, POLL is 160 bits, resource request is 32 bits, and sensing report is 32+sensing channel*4*8 bits.
The following is the residual WIFI parameter setting:
Fig22. 802.11 parameter setting
37
6.1 Maximum Coexistence SUs
We first give the value of WIFI user, timeout, and average packet size, and find the maximum coexistence SUs. The analysis result show coexistence number decrease with increasing packet size and increase with increasing timeout.
Fig23. Maximum coexistence SUs with different packet size and timeout Then, we want to know the coexistence SUs with different access scheme of WIFI users. The analysis result show the coexistence SUs in RTS/CTS scheme is smaller than basic scheme, but is large than basic scheme with lots of WIFI users.
Fig24. Maximum coexistence SUs with different access scheme
38
6.2 WIFI Throughput with SUs
We want to know when there exist some SUs in the WIFI network, the impact to the WIFI throughput. We first show the traditional throughput that doesn’t exist SUs.
Then, calculate the throughput when give the different value of WIFI users, SUs, and timeout. The analysis result show the impact to WIFI throughput increase when decrease timeout and increase coexistence SUs. But the impact is not obvious to WIFI throughput.
Basic Access
M traditional SU=50,t=1 SU=50,t=10 SU=100,t=1 SU=100,t=10 10 5.82198 5.82154 5.82194 5.8211 5.8219 30 5.07402 5.07388 5.074 5.07374 5.07399 50 4.6889 4.68882 4.6889 4.68874 4.68889
Fig25. WIFI throughput with coexistence SUs by basic access
The throughput impact in RTS/CTS is also not obvious to WIFI throughput RTS/CTS Access
M traditional SU=50,t=1 SU=50,t=10 SU=100,t=1 SU=100,t=10 10 4.70566 4.7053 4.70559 4.7053 4.70559 30 4.48614 4.48602 4.48612 4.48602 4.48612 50 4.35001 4.34994 4.35 4.34994 4.35
Fig26. WIFI throughput with coexistence SUs by RTS/CTS access
Through the analysis result, we know the impact of coexistence SUs is little for WIFI throughput no matter we adopt basic access scheme or RTS/CTS access scheme.
39
6.3
System Throughput
We calculate the system throughput in this section. Give M=20, K=20, timeout=10s, packet size =1500 bytes. The report error probability s =0.1, and we want to know the impact to system throughput when we give different value of WIFI throughput impact. When the throughput threshold increases, we can poll more SUs for collecting sensing report. That is, we just need fewer sensing time of QP, and increase the CR throughput. When we increase the impact threshold, we can find the suitable PCF duration for sensing report and request issue.
Fig27. System throughput1
40
Then, Give M=20, K=20, timeout=10s, packet size =1500 bytes, and impact threshold is 0.1. We want to know the impact for system throughput when the report error probability increases. Through the analysis result, we can see that we need to poll more SUs to acquire the require number of sensing report and it would cause more PCF duration. That is, decrease the WIFI throughput. In the other hands, we also need more sensing time of QP for sensing report and decrease the CR throughput.
Fig29. System throughput2
41
Then, Give M=20, timeout=10s, packet size =1500 bytes, impact threshold is 0.1, and report error probability 0.1. The sensing channel is 8 and available channel is 1.
We want to know the impact to throughput with different SU number. When SU number increases, more SUs could sense channel at the same time. It needs fewer channel sensing time, and increase the CR throughput.
Fig31. System throughput with different SU number
Fig32. WIFI throughput with different SU number
SU 10 15 20 25 30 35 40
CP(s) 0.05615 0.06707 0.089 0.13288 0.13288 0.13288 0.13288 RP(s) 0 0.01152 0.01347 0.01152 0.01347 0.01541 0.01736
QP(s) 0.05 0.04 0.03 0.02 0.02 0.02 0.02
Fig33. The duration of every period with different SU number
0
42
Then, give K=20, timeout=10s, packet size =1500 bytes, impact threshold is 0.1, and report error probability 0.01. The sensing channel is 8, and PCF duration is 0.011339s. We use the ns-2 simulator to validate the WIFI analysis result. The simulation result doesn’t match the analysis result because the analysis result doesn’t consider the packet transmit error probability and other control message. The
Then, give K=20, timeout=10s, packet size =1500 bytes, impact threshold is 0.1, and report error probability 0.01. The sensing channel is 8, and PCF duration is 0.011339s. We use the ns-2 simulator to validate the WIFI analysis result. The simulation result doesn’t match the analysis result because the analysis result doesn’t consider the packet transmit error probability and other control message. The