Cross-layer Scheduling and Resource Allocation in Wireless ...skrishna/nitc-fdp11.pdf · Dynamic...
Transcript of Cross-layer Scheduling and Resource Allocation in Wireless ...skrishna/nitc-fdp11.pdf · Dynamic...
Cross-layer Scheduling and Resource Allocation inWireless Communication Systems
Srikrishna Bhashyam
Department of Electrical EngineeringIndian Institute of Technology Madras
21 June 2011
Srikrishna Bhashyam (IIT Madras) 21 June 2011 1 / 44
Cellular Systems
Time-varying channel
Resource sharing – Interference constraints
Srikrishna Bhashyam (IIT Madras) 21 June 2011 2 / 44
Downlink Resource Allocation Problem
Channel information
User 1
User 2
User K
TrafficBasestation
Physical resources: power and bandwidthTotal transmit power constraintMaximize system throughputFairness or Quality of Service (QoS) constraints
Srikrishna Bhashyam (IIT Madras) 21 June 2011 3 / 44
Dynamic Resource Allocation
User 1 User 2 User 3
Periodic
reallocation of
resources
Resources: Time, Bandwidth, Power
Adaptation to channel and traffic conditionsDynamic resource allocation
I Reallocation period of the order of a millisecond
Srikrishna Bhashyam (IIT Madras) 21 June 2011 4 / 44
Adapting to the Channel
Srikrishna Bhashyam (IIT Madras) 21 June 2011 5 / 44
Adapting to the Channel: Maximizing Capacity
User 1
User 2
User K
Basestation
Select theuser with
best channel
Channel K
Channel 2
Channel 1
Infinite backlog assumption
All power and bandwidth resources to one user
User with best achievable rate chosen:
i = arg maxk
Rk ,
where Rk is the rate that can be supported by user k .
Srikrishna Bhashyam (IIT Madras) 21 June 2011 6 / 44
Maximizing Capacity: Parallel Channels
User 1
User 2
User K
Basestation
For each
best channeluser with
Select the
parallelchannel
Parallel Channels to each user
Bandwidth resources split to achieve parallel channels
For each channel n, user with best channel conditions chosen:
in = arg maxk
Rk,n.
Water-filling power allocation
Srikrishna Bhashyam (IIT Madras) 21 June 2011 7 / 44
Fairness
Proportional Fairness
i = arg maxk
Rk
Rk,av,
where Rk,av is the average rate that can be supported by user k .
max∑
k log (Tk),
where Tk is the average long-term throughput of user k .
Srikrishna Bhashyam (IIT Madras) 21 June 2011 8 / 44
Parallel Channels: OFDM
Power
User 1
User 2
User 3Subc
arri
ers
Available resources:I SubcarriersI Transmit power
Channel is frequency-selective ⇒ subcarriers not identical.
Srikrishna Bhashyam (IIT Madras) 21 June 2011 9 / 44
Fairness: Joint Subchannel and Power Allocation
Proportional rate subcarrier allocation [Rhee]
Proportional rate subcarrier allocation + power optimization [Shen]
Joint subcarrier and power allocation
Srikrishna Bhashyam (IIT Madras) 21 June 2011 10 / 44
Fairness: Joint Subchannel and Power Allocation
power
Split power equally amongst subcarriers
Start
Checkif all subcarriers
are allocated
Allocate a subcarrier to a user
Start
Split power equally amongst subcarriers
Allocate all subcarriers to users
Update each user’squeue
No
Yes
End
Optimize powerallocation with
Update each user’squeue
End
Optimize power
allocation with
power
Srikrishna Bhashyam (IIT Madras) 21 June 2011 11 / 44
Gradient Algorithm
Stolyar (2005)
General utility functions
Multiuser scheduling at the same time
PF is a special case
Srikrishna Bhashyam (IIT Madras) 21 June 2011 12 / 44
Adapting to the Channel and Traffic
Srikrishna Bhashyam (IIT Madras) 21 June 2011 13 / 44
Adapting to the Channel and Traffic
Queues for each user
Servers
Users
Time-varying connectivity
Multi-Queue Multi-Server Model for each time slot
Server: Subcarrier/Group of subcarriers/Spreading code
Srikrishna Bhashyam (IIT Madras) 21 June 2011 14 / 44
Resource Allocation/Cross-layer Scheduling Goals
Scheduling GoalsI Stability and throughput optimality
F Stability: Average queue length finite
Arrival rate of user 1
Arr
ival
rat
e of
use
r 2
Stability region of policy 1
Stability region of policy 2
Stability region of throughput optimal policy
I Packet delay constraintsI Fairness
Srikrishna Bhashyam (IIT Madras) 21 June 2011 15 / 44
Stability in a general wireless network
[Tassiulas et al 1992, Georgiadis et al 2006]I Dynamic backpressure policy
Destination node
b1 r14
b2
b3r34
b1 b3
(b2 − b1)r21
(b1 − b3)r13
b2r24
Interference model: Only certain links can be activated simultaneously
Scheduling problem: Which links will you activate?
Solution: Activate those links such that the sum of their weights ismaximum.
Srikrishna Bhashyam (IIT Madras) 21 June 2011 16 / 44
Dynamic back-pressure policy for our settingMax-Weight Scheduling
Users Serversb1(t)
b2(t)
b1C11
b2C21
b1C
12
b2C22
Only one link per server to be activated. Which links to activate?
Solution:I Make the servers as destination nodes.I Assign the weights for each link as in back-pressure policy.I Activate those links such that the sum of their weights is maximum.
max∑
k
bnCnk
bn: Backlog of user n, Cnk : Capacity of user n on server k
Srikrishna Bhashyam (IIT Madras) 21 June 2011 17 / 44
Two Throughput Optimal Policies
Users Serversb1
b2
b3
b1C11
b1C12
b3C32
b2C21
b 3C 31
b2C22
Policy 1: Max-Weight Scheduling
Policy 2: Improving delay performanceI Update queue information after each server is scheduled
Srikrishna Bhashyam (IIT Madras) 21 June 2011 18 / 44
Joint Server and Power Allocation
Finite number of power levelsI Max-weight scheduling
Joint subcarrier and power allocationI Joint optimizationI Sub-optimal solutions
Srikrishna Bhashyam (IIT Madras) 21 June 2011 19 / 44
Results: Max. Arrival Rate vs. Transmit Power
4 4.5 5 5.5 6 6.5 7 7.5 82.5
3
3.5
4
4.5
5
5.5
6
6.5
7
7.5
Ptotal in dBW
Arriv
al ra
te (M
bps)
CAO+FPACAO+FPA+PAOCAO+JSPAMLWDFCAQA+FPACAQA+JSPAHomogenous rate users
Max. arrival rate for less than 0.5% packets droppedSrikrishna Bhashyam (IIT Madras) 21 June 2011 20 / 44
Results: Delay Performance
5 10 15 20 25 30 35 40
10−1
100
101
delay (time slots)
P(de
lay>
x)
CAO+FPACAO+FPA+PAOCAO+JSPAMLWDFCAQA+FPACAQA+JSPA
Ptotal = 8dBW Arrival rate = 3 MbpsHomogenous rate users
Best and worst delay performance among users plotted
Srikrishna Bhashyam (IIT Madras) 21 June 2011 21 / 44
Fairness and Utility Maximization
Arrival rate vector outside stability regionI Support a fraction of the trafficI Optimize utility based on long term throughputI Flow control to get stabilizable rates + stabilizing policyI Fairness based on choice of utility function
F Proportional fairness
Srikrishna Bhashyam (IIT Madras) 21 June 2011 22 / 44
Fairness and Utility Maximization
Flow control + stabilizing policy
Maximize utility subject to stability
max{rk}
∑k
[Vfk(rk)− bk rk ]
Srikrishna Bhashyam (IIT Madras) 21 June 2011 23 / 44
Adapting with Partial Information
Srikrishna Bhashyam (IIT Madras) 21 June 2011 24 / 44
Using Delayed Information
Second Interval Third IntervalFirst Interval
Slot
T
b(T − 1), C (T − 1) b(2T − 1), C (2T − 1)
T T
Time-slots are grouped into intervals
Channel and queue information available only once in T slots
Srikrishna Bhashyam (IIT Madras) 21 June 2011 25 / 44
Channel model
USER 1’s CHANNEL USER 2’s CHANNEL
1 2 3 4 5 6
0
1
2
3
654321
0
1
2
3
SERVERS SERVERS
PACKETS PACKETS
Cnk : channel capacity of user n on server k.
Cnk ∈ {0, 1, 2, 3}.
Srikrishna Bhashyam (IIT Madras) 21 June 2011 26 / 44
Loss model
0
Packets sent
Rate
Capacity
Rnk
Rnk <= Cnk Rnk > Cnk
Rnk : number of packets user n transmits on server k .
Cnk(lT − 1): channel information available at the start of l th interval.
Srikrishna Bhashyam (IIT Madras) 21 June 2011 27 / 44
Scheduling with infrequent measurements
Retain throughput optimality of dynamic backpressure policy
Two policies: Policy 1 and Policy 2
Comparison with KLS policy [Kar et al 2007]
Srikrishna Bhashyam (IIT Madras) 21 June 2011 28 / 44
Policy 1 & Policy 2Users Servers
b1
b2
b3
b1C11
b1 C12
b3C32
b 3C 31
b2C21
b2C22
Define Cnk = max E [Tnk(t)|Cnk(lT − 1)]
= maxr
r Pr{r ≤ Cnk | Cnk(lT − 1)}
Policy 1 is the dynamic back pressure policy for our setting
Assignment changes every slot
Policy 2: Update queue information after each server is scheduled
Srikrishna Bhashyam (IIT Madras) 21 June 2011 29 / 44
KLS Policy
Users Servers
b31C31
b11
b21
b31
b12
b22
b32
b32C32
b12 C
12
b11 C
11
b21C21
b22C22
Virtual queue for each user-server pair
Define Cnk(lT ) as
1
TE
(l+1)T−1∑t=lT
Cnk(t)∣∣∣Cnk(lT − 1)
Assignment changes once in T slots
Srikrishna Bhashyam (IIT Madras) 21 June 2011 30 / 44
Simulation setup
Truncated Poisson arrivals
128 users and 16 servers
Markov fading channel with probability transition matrix
Backlog and delay are used as metrics for comparison
Simulations for both symmetric and asymmetric arrivalsI Symmetric case shown here
Srikrishna Bhashyam (IIT Madras) 21 June 2011 31 / 44
Average backlog comparison: Slow fading, T = 8
5 10 15 20 25 30 35 40 45 500
100
200
300
400
500
600
700
Net arrival rate
Ave
rage
bac
klog
in p
acke
ts/ti
mes
lot/u
ser
KLS PolicyPolicy 1Policy 2
All the policies have similar stability region.
Srikrishna Bhashyam (IIT Madras) 21 June 2011 32 / 44
Average backlog comparison for low traffic
5 10 15 20 25 300
20
40
60
80
100
120
140
160
180
Net arrival rate
Ave
rage
bac
klog
in p
acke
ts/ti
me−
slot
/use
r
KLS PolicyPolicy 1Policy 2
At low traffic, proposed policies outperform KLS policy.
Srikrishna Bhashyam (IIT Madras) 21 June 2011 33 / 44
Delay comparison
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.1810
−2
10−1
100
delay ’d’ in seconds
Pro
babi
lity
that
del
ay is
> ’d
’
min(delay) policy 2max(delay) policy 2min(delay) policy 1max(delay) policy 1min(delay) KLS policymax(delay) KLS policy
Net arrival rate = 25.6, T = 4
Srikrishna Bhashyam (IIT Madras) 21 June 2011 34 / 44
Average backlog comparison vs T for Policy 2
0 10 20 30 40 50 60 700
100
200
300
400
500
600
700
Net arrival rate
Ave
rage
bac
klog
in p
acke
ts/ti
mes
lot/u
ser
T=1T=2T=4T=8T=15
Srikrishna Bhashyam (IIT Madras) 21 June 2011 35 / 44
Average backlog comparison for different policies
5 10 15 20 25 30 35 40 450
100
200
300
400
500
600
700
Net arrival rate
Ave
rage
bac
klog
in p
acke
ts/ti
mes
lot/u
ser
KLS PolicyPolicy 1Policy 2
Srikrishna Bhashyam (IIT Madras) 21 June 2011 36 / 44
Comparison of stability regions: Fast fading
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.80
100
200
300
400
500
600
700
Net arrival rate
Ave
rage
bac
klog
in p
acke
ts/ti
mes
lot/u
ser
Policy 1KLS PolicyPolicy 2
2 queues, 1 server, T = 2, states are {0, 1}
Probability transition matrix:
[δ 1− δ
1− δ δ
], δ = 0.1
Srikrishna Bhashyam (IIT Madras) 21 June 2011 37 / 44
Possible Extensions
Srikrishna Bhashyam (IIT Madras) 21 June 2011 38 / 44
More Physical Layer Options
Multiple antennas
Power allocation across resources (servers)
Interference processing vs. Interference avoidance
Multi-cell scenario: Centralized vs. Distributed methods
Srikrishna Bhashyam (IIT Madras) 21 June 2011 39 / 44
Approximate Solutions
Lower complexity/approximate solutions to optimization problem
Appropriate reduction search space of physical layer modes
Srikrishna Bhashyam (IIT Madras) 21 June 2011 40 / 44
Summary
Adapting to the channel
Adapting to the channel and trafficI Max-weight Scheduling
Adapting to partial informationI Conditional expected rate
Possible extensionsI Approximate lower complexity solutionsI Appropriate choice of physical layer modes
Srikrishna Bhashyam (IIT Madras) 21 June 2011 41 / 44
References
R. Knopp, P. Humblet, “Information Capacity and power control in single cell multiuser communications,” in Proc. IEEE
ICC, Seattle, WA, vol. 1, pp. 331-335, June 1995.
D. N. C. Tse, “Optimal power allocation over parallel Gaussian channels,” in Proc. IEEE ISIT, Ulm, Germany, pp. 27,
June 1997.
E. F. Chaponniere, P. Black, J. M. Holtzman, and D. Tse, “Transmitter directed multiple receiver system using path
diversity to equitably maximize throughput,” U. S. Patent No. 6449490, September 2002.
P. Viswanath, D. N. C. Tse, R. Laroia, “Opportunistic beamforming using dumb antennas,” IEEE Transactions on
Information Theory, vol. 48, no. 6, pp. 1277-1294, June 2002.
C. Y. Wong, R. S. Cheng, K. B. Letaief, R. D. Murch, “Multiuser OFDM with Adaptive Subcarrier, Bit, and Power
Allocation”, IEEE Journal on Selected Areas in Communications, vol. 17, no. 10, pp. 1747-1758, October 1999.
J. Jang, K. B. Lee, “Transmit power adaptation for multiuser OFDM systems”, IEEE Journal on Selected Areas in
Communications, vol. 21, no. 2, pp. 171-178, February 2003.
W. Rhee, J. M. Cioffi, “Increase in Capacity of Multiuser OFDM System Using Dynamic Subchannel Allocation”,
Proceedings of the 51st IEEE Vehicular Technology Conference, Tokyo, vol. 2, pp. 1085-1089, Spring 2000.
Z. Shen, J. G. Andrews, B. L. Evans, “Adaptive Resource Allocation in Multiuser OFDM Systems with Proportional Rate
Constraints”, IEEE Transactions on Wireless Communications, vol. 4, no. 6, pp. 2726-2737, November 2005.
C. Mohanram, S. Bhashyam, “A sub-optimal joint subcarrier and power allocation algorithm,” IEEE Communications
Letters, vol. 9, no. 8, pp. 685-687, August 2005.
Srikrishna Bhashyam (IIT Madras) 21 June 2011 42 / 44
References
L. Tassiulas, A. Ephremides, “Stability properties of constrained queueing systems and scheduling for maximum
throughput in multihop radio networks,”IEEE Transactions on Automatic Control, vol. 37, no. 12, pp. 1936-1949,December 1992.
L. Georgiadis, M. J. Neely, L. Tassiulas, “Resource allocation and cross-layer control in wireless networks,” Foundations
and Trends in Networking, vol. 1, no. 1, pp. 1-144, 2006.
M. Andrews, K. Kumaran, K. Ramanan, A. L. Stolyar, R. Vijayakumar, P. Whiting, “Providing quality of service over a
shared wireless link,” IEEE Communications Magazine, vol. 39, no. 2, pp. 150-154, Feb 2001.
A. L. Stolyar, “On the asymptotic optimality of the gradient scheduling for multi-user throughput allocation,” Operations
Research, vol. 53, no. 1, pp. 12-25, 2005.
S. Kittipiyakul, T. Javidi, “Resource allocation in OFDMA with time-varying channel and bursty arrivals,” IEEE
Communication letters, vol. 11, no. 9, September 2007.
C. Mohanram, S. Bhashyam, “Joint subcarrier and power allocation in channel-aware queue-aware scheduling for
multiuser OFDM,” IEEE Transactions on Wireless Communications, vol. 6, no. 9, September 2007.
K. Kar, X. Luo, S. Sarkar, “Throughput-optimal scheduling in multichannel access point networks under infrequent
channel measurements,” IEEE Transactions on Wireless Communications, vol. 7, no. 7, pp. 2619-2629, July 2008.
C. Manikandan, S. Bhashyam, R. Sundaresan, “Cross-layer scheduling with infrequent channel and queue
measurements,” IEEE Transactions on Wireless Communications, vol. 8, no. 12, pp. 5737-5742, December 2009.
P. Chaporkar, K. Kar, X. Luo, S. Sarkar, “Throughput and Fairness Guarantees through maximal scheduling in wireless
networks,” IEEE Transactions on Information Theory, vol. 54, no. 2, pp. 572-594, February 2008.
Srikrishna Bhashyam (IIT Madras) 21 June 2011 43 / 44
Acknowledgements
Rajesh Sundaresan
Chandrashekar Mohanram
C. Manikandan
Parimal Parag
Department of Science and Technology
Srikrishna Bhashyam (IIT Madras) 21 June 2011 44 / 44