Title: Wireless packet scheduling in an integrated CDMA system using channel status information
1Wireless packet scheduling in an integrated CDMA
system using channel status information
- AuthorSeung Sik Choi and Dong Ho Cho
- SourceComputer Communications, Volume 27,
Number 9, June 2004, Pages 890-897 - Reporter Tsang-Yuan Tsai
- Date2005/1/15
2Outline
- Introduction
- System Model with channel status
- Implementation issues
- Virtual time implementation
- Credit table implementation
- Weighted Fair Queueing with State control (WFQS)
- Simulation Results
- Conclusions
3Outline
- Introduction
- System Model with channel status
- Implementation issues
- Virtual time implementation
- Credit table implementation
- Weighted Fair Queueing with State control (WFQS)
- Simulation Results
- Conclusions
4CDMA system Characteristics
- Interference limited
- Different Bit Error Rate (BER) values for
multimedia services - Multiple sessions can be served simultaneously
using different pseudonoise (PN) codes.
5Types of CDMA
6Throughput maximization strategy
- Delay-sensitive voice users are allocated first,
and then residual resources are allocated to data
users. - Nv is the no of voices users and Nd is the no of
data users. N Nv Nd
7Outline
- Introduction
- System Model with channel status
- Implementation issues
- Virtual time implementation
- Credit table implementation
- Weighted Fair Queueing with State control (WFQS)
- Simulation Results
- Conclusions
8Base station scheduler
- Virtual Packet Generator
- makes virtual packet according to the size of
frame and rate requirement - Power Resource Allocation Find valid spreading
codes and determined power index
9Channel state model using a two-state Markov
chain (1/3)
The steady-state probability Pe that a frame
error occurs can be given by
10Outline
- Introduction
- System Model with channel status
- Implementation issues
- Virtual time implementation
- Credit table implementation
- Weighted Fair Queueing with State control (WFQS)
- Simulation Results
- Conclusions
11Virtual time implement (1/3)
- The virtual time v(t) is defined
- Consider the kth virtual packet of a given
session i with a length of Lki bits. Suppose that
this virtual packet arrives at time aki, is
serviced at time ski, and is ended at dki. Then,
the virtual finishing time can be obtained by
where max(Fik-1, v(aki)) Fik-1 means no
backlog max(Fik-1i, v(aik)) Fik-1
means backlog
12Virtual time implement (2/3)
- Provided
- two sessions (blue,red)
- service rate 400bps is the same in the blue and
red sessions - Packet length 100bytes 800bits
9
2
1
9
1
F1blue 800/400 1 3 F1red 800/400 3 5
13Virtual time implement (3/3)
9
2
1
9
1
2
F2blue 800/400 3 5 F2red 800/400 6 8
14Outline
- Introduction
- System Model with channel status
- Implementation issues
- Virtual time implementation
- Credit table implementation
- Weighted Fair Queueing with State control (WFQS)
- Simulation Results
- Conclusions
15Credit table (1/2)
- Let Ci(tn) be the credit value of the ith user in
the time tn -
-
- where Sdi(tn,tn-1) is power resource of
data session i - during (tn,tn-1 and Fi is service shared
for ith user
16Credit table (2/2)
- Let Vi(tn) be the normalized credit value in the
nth frame -
- where nf is the number of frames during the call.
- The lowest Vi(n) value user is allocated first.
17 Example Assumption initial value
- Three users user 1, user2 and user3
- Service share F12R, F2R and F3R
- where R is the data length (bit/frame) to be
transmitted in a frame with 128 spreading gain. - Data rate Chip rate (fixed) / Spreading gain
18Example - User burst traffic
19Example Credit table initial
20Example Frame1
- User1
- C1(1) 0 2R/2R 1
- V1(1) 1/1 1
21Example Frame2
- User3( V3(1) is the lowest )
- C3(2) 0 R/R 1
- V3(2) 1/1 1
- User1
- C1(2) 1 R/2R 3/2
- V1(2) (3/2)/2 3/4
22Example Frame3
- User2( V2(2) is the lowest )
- C2(3) 0 R/R 1
- V2(3) 1/1 1
- User1(V1(2) lt V3(2)
- C1(3) 3/2 R/2R 2
- V1(3) 2/3
23Example Frame4
- User3( V3(3) is the lowest )
- C3(4) 1 R/R 2
- V3(4) 2/3
- User1(V1(3) lt V2(3)
- C1(4) 3/2 R/2R 2
- V1(4) 2/3
24Example Frame5
- User2( V2(4) is the lowest )
- C2(5) 1 R/R 2
- V2(5) 2/3
- User1(V1(4) lt V2(4)
- C1(5) 5/2 R/2R 3
- V1(5) 3/5
- Until frame 5,
- user 1,2 and 3 transmit 6R,
- 2R, and 2R
25Outline
- Introduction
- System Model with channel status
- Implementation issues
- Virtual time implementation
- Credit table implementation
- Weighted Fair Queueing with State control (WFQS)
- Simulation Results
- Conclusions
26Weighted Fair Queueing with State control (1/2)
Radio resource enough ?
Based on two- state Markov chain
27Weighted Fair Queueing with State control (2/2)
Fik small first
Credit table
28Outline
- Introduction
- System Model with channel status
- Implementation issues
- Virtual time implementation
- Credit table implementation
- Weighted Fair Queueing with State control (WFQS)
- Simulation Results
- Conclusions
29Simulation assumption
- Voice source packets MMPP (Markov Modulated
Poisson Process) - Data source packets Poisson process
- Frame size 20msec
- Transmission rate of data user 8kbps to 64kbps
- Burst error p 0.95 and q varies from 0.5 to
0.7 - Average length of the bad state
- 44 ms in q 0.5
- 66 ms in q 0.7
30Delay in bad state channel consideration (1/2)
- Delay of each data user when a bad state channel
is given to - data mobile 4 (Nv16)
31Throughput in bad state channel consideration
(2/2)
- Throughput of each data user when a bad state
channel is - given to data mobile 4 (Nv16)
32Simulation Result (1/2)
Delay of data users vs number of voice users (p
0.98)
33Simulation Result (2/2)
Throughput of data users vs number of voice users
(p 0.98)
34Conclusions
- Weighted fair queueing for data service in a
voice/data integrated CDMA - Throughput can be increased compared with a
conventional WFQ scheme - In reliable network the WFQS will get better
effective. -