a theory of qos for wireless
DESCRIPTION
A Theory of QoS for Wireless. I-Hong Hou Vivek Borkar P.R. Kumar. University of Illinois, Urbana-Champaign. Background: Wireless Networks. There will be increasing use of wireless networks for serving traffic with QoS constraints: VoIP Video Streaming Real-time Monitoring - PowerPoint PPT PresentationTRANSCRIPT
A Theory of QoS for Wireless
I-Hong Hou
Vivek Borkar
P.R. Kumar
University of Illinois,
Urbana-Champaign
Background: Wireless Networks
There will be increasing use of wireless networks for serving traffic with QoS constraints:
– VoIP
– Video Streaming
– Real-time Monitoring
– Networked Control
1/19
Challenges
How to formulate a relevant and tractable framework for QoS?
Relevant – Jointly deal with three criteria:– Deadlines– Delivery ratios– Channel unreliabilities
Tractable – Provide solutions for QoS support:– Admission control policies for flows– Scheduling policies for packets
2/19
Client-Server Model
A system with N wireless clients and one AP
Time is slotted
One packet transmission in each slot– successful with probability pn
– failed with probability 1-pn
Non-homogeneous link qualities– p1,p2,…,pN may be different
AP
12
3
N
slot length = transmission duration
p1 p2
p3
pN
3/19
Protocol for Operation
AP indicates which client should transmit in each time slot
Uplink– Poll (ex. CF-POLL in 802.11 PCF)DataNo need for ACKpn = Prob( both Poll/Data are delivered)
DownlinkDataACKpn = Prob( both Data/ACK are delivered)
AP
n
CF-POLL
4/19
Protocol for Operation
AP indicates which client should transmit in each time slot
Uplink– Poll (ex. CF-POLL in 802.11 PCF)– DataNo need for ACKpn = Prob( both Poll/Data are delivered)
DownlinkDataACKpn = Prob( both Data/ACK are delivered)
AP
nData
4/19
Protocol for Operation
AP indicates which client should transmit in each time slot
Uplink– Poll (ex. CF-POLL in 802.11 PCF)– Data– No need for ACK– pn = Prob( both Poll/Data are delivered)
DownlinkDataACKpn = Prob( both Data/ACK are delivered)
AP
n
4/19
Protocol for Operation
AP indicates which client should transmit in each time slot
Uplink– Poll (ex. CF-POLL in 802.11 PCF)– Data– No need for ACK– pn = Prob( both Poll/Data are delivered)
Downlink– DataACKpn = Prob( both Data/ACK are delivered)
AP
n
Data
4/19
Protocol for Operation
AP indicates which client should transmit in each time slot
Uplink– Poll (ex. CF-POLL in 802.11 PCF)– Data– No need for ACK– pn = Prob( both Poll/Data are delivered)
Downlink– Data– ACK– pn = Prob( both Data/ACK are delivered)
AP
nACK
4/19
QoS Model
time slot
Timeline of the system
5/19
QoS Model
Clients generate packets with fixed period τDeadline = PeriodPackets expire and are dropped if not delivered by the
deadlineDelay of successfully delivered packet is at most τDelivery ratio of packets that meet deadline for client n
should be at least qn
arrival arrival arrival arrivalτ
5/19
QoS Model
Clients generate packets with fixed period τ Deadline = Period Packets expire and are dropped if not delivered by the
deadlineDelay of successfully delivered packet is at most τDelivery ratio of packets that meet deadline for client n
should be at least qn
arrival deadline
5/19
QoS Model
Clients generate packets with fixed period τ Deadline = Period Packets expire and are dropped if not delivered by the
deadline Delay of successfully delivered packet is at most τDelivery ratio of packets that meet deadline for client n
should be at least qn
arrival deadlineτ
5/19
QoS Model
Clients generate packets with fixed period τ Deadline = Period Packets expire and are dropped if not delivered by the
deadline Delay of successfully delivered packet is at most τ Delivery ratio of packets that meet deadline for client n
should be at least qn
#delivered packets in periodsliminf , a.s., nn
K
Kq
K
delivered delivered
dropped
5/19
Problem Formulation
Admission control– Decide whether a set of clients is feasible
Scheduling policy– Design a policy that fulfills every feasible set of
clients
6/19
The proportion of time slots needed for client n is
Work Load
1 nn
n
qw
p τ
7/19
The proportion of time slots needed for client n is
Work Load
1 nn
n
qw
p τ
expected number of time slots needed for a successful transmission
7/19
The proportion of time slots needed for client n is
Work Load
1 nn
n
qw
p τ
number of required successful transmissions in a period
7/19
The proportion of time slots needed for client n is
Work Load
1 nn
n
qw
p τ
normalize by period length
7/19
The proportion of time slots needed for client n is
We call wn the “work load”
Work Load
1 nn
n
qw
p τ
7/19
Necessary Condition for Feasibility of QoS Requirements
Necessary condition from classical queuing theory:
But the condition is not sufficient Packet drops by deadline misses cause more
idleness than in queuing theory
11
N
nnw
AP
1 2
S XS X
forced idleness
8/19
Let I be the expected proportion of idle time
Stronger necessary condition
Sufficient?
Stronger Necessary Condition
1
1N
nn
w I
Still NO!
9/19
Even Stronger Necessary Condition
Let IS = Expected proportion of the idle time when the set of clients is S– IS decreases as S increases
Theorem: the condition is both necessary and sufficient
1, {1,2,..., }n Sn S
w I S N
10/19
Largest Debt First Scheduling Policies
Give higher priority to client with higher “debt”
AP
1 2
3
SF FF
SF
11/19
Two Definitions of Debt
The time debt of client n – time debt = wn – actual proportion of transmission time
given to client n
The weighted delivery debt of client n– weighted delivery debt = (qn – actual delivery ratio)/pn
Theorem: Both largest debt first policies fulfill every feasible set of clients
– Feasibility Optimal Policies
12/19
Tractable Policy for Admission Control
Admission control consists of determining feasibility
Apparently, we need to check:
2N tests, so computationally complex
Theorem: Can be made into N tests– Polynomial time algorithm for admission control– Order clients by qn
1, {1,2,..., }n Sn S
w I S N
13/19
Simulation Setup
Implement on IEEE 802.11 Point Coordination Function (PCF)
Application: VoIP standard
Deadline miss ratio (DMR) = shortfall in delivery ratio
64 kbp data rate 20 ms period
160 Byte packet 11 Mb/s transmission rate
610 μs time slot 32 time slots in a period
14/19
Evaluated Four Policies
DCF
PCF with randomly assigned priorities
Time debt first policy
Weighted-delivery debt first policy
15/19
Test at Edge of Feasibility
Two groups of clients:– Group A requires 99% delivery ratio– Group B requires 80% delivery ratio– The nth client in each group has (60+n)% channel
reliability Feasible set: 6 group A clients and 5 group B
clients Infeasible set: 6 group A clients and 6 group
B clients
16/19
Results for a Feasible Set
0
1
2
3
4
5
6
0 1 2 3 4 5
Time (sec)
DM
R
Weighted-delivery
17/19
Results for a Feasible Set
0
1
2
3
4
5
6
0 1 2 3 4 5
Time (sec)
DM
R
Weighted-deliveryTime-based
fulfilled
17/19
Results for a Feasible Set
0
1
2
3
4
5
6
0 1 2 3 4 5
Time (sec)
DM
R
Weighted-deliveryTime-based
Random
17/19
Results for a Feasible Set
0
1
2
3
4
5
6
0 1 2 3 4 5
Time (sec)
DM
R
Weighted-deliveryTime-based
Random
DCF
17/19
Results for an Infeasible Set
0
1
2
3
4
5
6
0 1 2 3 4 5
Time (sec)
DM
R
Weighted-delivery
18/19
Results for an Infeasible Set
0
1
2
3
4
5
6
0 1 2 3 4 5
Time (sec)
DM
R
Weighted-deliveryTime-based
small DMR
18/19
Results for an Infeasible Set
0
1
2
3
4
5
6
0 1 2 3 4 5
Time (sec)
DM
R
Weighted-deliveryTime-based
Random
18/19
Results for an Infeasible Set
0
1
2
3
4
5
6
0 1 2 3 4 5
Time (sec)
DM
R
Weighted-deliveryTime-based
Random
DCF
18/19
Conclusion
Formulate a framework for QoS that deal with deadlines, delivery ratios, and channel unreliabilities
Characterize when QoS is feasible
Provide efficient scheduling policy
Provide efficient admission control policy
Address implementation issues
19/19
Backup Slides
An example:– Two clients, τ = 3– p1=p2=0.5– q1=0.876, q2=0.45– w1=1.76/3, w2=0.3– I{1}=I{2}=1.25/3, I{1,2}=0.25/3
w1+I{1}=3.01/3 > 1 However, w1+w2+I{1,2}=2.91/3 < 1