Nov 2009
doc.: IEEE 802.11-09/1263r1
[Preliminary Simulation Results on Power Saving]
Date: 2009-11-19
Authors:
Name
Affiliations
Address
Phone
email
Zhuo Chen
Rutgers
University
+1-848565-8125
[email protected]
Chunhui (Allan)
Zhu
Samsung
Electronics
+1-408544-5667
[email protected]
Shweta Jain
Rutgers
University
+1-631455-6193
[email protected]
Youngsoo Kim
Samsung
Electronics
671 Route 1 South,
North Brunswick, NJ
08901
75 W. Plumeria Dr.
San Jose, CA 95131
USA
671 Route 1 South,
North Brunswick, NJ
08901
Mt. 14-1 Nongseo-Ri,
Giheung-Eup,
Yongin-Si, GyeonggiDo, Korea 449-712
+82-31280-9614
[email protected]
Submission
Slide 1
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
Abstract
• This presentation reports the preliminary results of our
simulations of existing power saving mechanisms.
– The scenario simulated is taken from IEEE 802 Doc 09/0161r02,
Usage Model 2d: Wireless Networking for Office.
– The performance and effect of APSD have been examined.
Submission
Slide 2
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
Simulation Scenario*
Pre-Conditions:
Office with people engaged in high quality/high revenue
services that involve video and voice interaction with client
and transferring large volumes of multimedia data. A single
AP serves 5 people. The office comprises 5 – 500 people.
Application:
Multiple applications run at the same time. High definition
compressed video uses something like an Blu-ray codec.
Voice is standard definition quality using a codec like G729.
Aggregate bandwidth requirement is 5 simultaneous video
streams per AP.
Voice requirements are: ~50Kbps, Jitter <30msec. Delay
<30msec. 1.0E-1 PER.
Traffic Conditions (per AP):
2 WLAN video streams
2 WVoIP streams
Up to 5 best effort data streams
The best effort data traffic can take up to
20% of the available bandwidth with
saturated offered load.
Use Case:
1.Users run different applications during the
day and may start each application at
different time.
2.A typical sequence is starting up a voice
call, adding video sending/receiving multiEnvironment:
media data and discussing this over the
Mostly not Line of sight within a single office.
voice/video link
People walking around the office.
3.The duration of such a use case is
There is potentially unmanageable interference from
typically one hour.
neighboring offices within 100 feet (horizontally) or adjacent
4.Up to three of these “sessions” per AP
floors (vertically) when in 2.4 / 5 GHz.
may be going on in parallel.
AP density is more than 1 AP per 40m X 40m
* This was taken from Doc 09/0161r02, Usage Model 2d: Wireless Networking for Office
Submission
Slide 3
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
Network Topology
Submission
Slide 4
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
Traffic Pattern
Source
Sink
STAs
Channel
Flow No.
Location Location
(Source/Sink)
Model
(meters) (meters)
Application
(Forward
Traffic /
Backward
Traffic)
Application
Rate
MSDU Size Max. Delay
Max.
Load (Mbps) Distribution
(B)
(ms)
PLR
(Forward / (Forward / (Forward / (Forward / (Forward /
Backward) Backward) Backward) Backward) Backward)
1
AP / STA1
(0,0)
(0,5)
C
VoIP / VoIP
0.008 / 0.008 UDP / UDP
20/ 20
30 / 30
5% / 5%
2
AP / STA2
(0,0)
(-10,-10)
C
VoIP / VoIP
0.008 / 0.008 UDP / UDP
20/ 20
30 / 30
5% / 5%
3
AP/ STA3
(0,0)
(5,0)
C
VoIP / VoIP
0.008 / 0.008 UDP / UDP
20/ 20
30 / 30
5% / 5%
4
STA3/ AP
(5,0)
(0,0)
C
Local file
transfer
300
INF
N/A
5
AP/ STA 4
(0,0)
(-7,7)
C
VoIP / VoIP
20/ 20
30 / 30
5% / 5%
Constant,
UDP /
Constant
UDP
1500 / 64
20 / 100
10^-7 / 10^-2
TCP
300
INF
N/A
Constant,
UDP /
Constant
UDP
1500 / 64
20 / 100
10^-7 / 10^-2
6
STA 4/ AP
(-7,7)
(0,0)
C
7
STA 5/ AP
(10,5)
(0,0)
C
8
STA 5/ AP
Submission
(10,5)
(0,0)
C
Max. 1Gbps
0.008 / 0.008 UDP / UDP
Blu-ray/
control channel 50.00 / 0.06
Local file
transfer
TCP
Max. 1Gbps
Blu-ray/
control channel 50.00 / 0.06
Slide 5
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
Simulation Parameters
Bit rate for DATA packets
Bit rate for RTS/CTS/ACK
aCWmin
aCWmax
PLCPDataRate
A Slot Time
SIFS
DIFS
PreambleLength
PLCPHeaderLength
MAC header
IP header
DATA packet
RTS
CTS, ACK
AC 3 for VoIP traffic
AC 2 for Video traffic
AC 0 for Video Best effort traffic
500 Mbps
6 Mbps
31
1024
6 Mbps
9 us
16 us
34 us
144 bits
48 bits
224 bits
160 bits
Payload size + MAC header + IP
header=Payload size+ 384 bits
160 bits
112 bits
PF 2, AIFS 2, CW_MIN 7, CW_MAX 15
PF 2, AIFS 2, CW_MIN 15, CW_MAX 31
PF 2, AIFS 7, CW_MIN 31, CW_MAX 1023
The simulation runs on ns2 platform.
Submission
Slide 6
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
Performance without Aggregation
Network Performance with APSD
Delay (ms)
Uplink
VoIP
Throughput (Kbps)
Downlink
Uplink
Packet Loss (%)
Downlink
Uplink
Downlink
0.361
10.622
8
8
0
0
Video
326.813
0.661
19,298
60
60.23
0
TCP
18.668
N/A
998.198
N/A
0
0
Network Performance without APSD
Delay (ms)
Uplink
VoIP
Throughput (Kbps)
Downlink
Uplink
Packet Loss (%)
Downlink
Uplink
Downlink
0.375
0.351
8
8
0
0
Video
304.989
0.619
19,907
60
59.06
0
TCP
22.137
N/A
1,068.31
N/A
0
0
Note the data here are average numbers over 4 VoIP sessions, 2 Video sessions and 2 TCP sessions.
Submission
Slide 7
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
802.11n Capacity with Video traffic
DIFS
34 µs
Contention
Window
67 µs
(min average)
PLCP
Header
28 µs
Video
Frame
SIFS
Ack
16 µs
32 µs
Time
1. Video Frame = (Data + IP header + MAC header)/DataRate
( 1500 +
20
+ 28 ) * 8
/ 500
= 24.768 µs
2. Require Bandwidth = Application Load * Frame Start Interval / Video Frame
50 * (34+67+28+24.768+16+32) / 24 = 420.315 Mbps
3. Number of Video uplink sessions = Bandwidth * Utilization / Require Bandwidth
500
*
0.8
/ 420.315
= 1.05
Reference: IEEE 802.11-07/2704r00, Efficiency of VoIP on 802.11n
Submission
Slide 8
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
Data Aggregation
• In order to support the given traffic load, Aggregated
MSDU is used.
– In the results after this slide, four video MSDUs are aggregated
into one A-MSDU.
– VoIP and TCP packets are not aggregated.
Submission
Slide 9
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
APSD Power Saving Effect
1.2
Receive Time
Transmit Time
Idle Time
1
Sleep Time
Percentages
0.8
0.6
0.4
0.2
0
Submission
STA11
2 2
STA
(VoIP only)
(VoIP only)
3 3
STA
STA Number
(VoIP, TCP)
Slide 10
4 4
STA
5 5
STA
(VoIP, Video) (VoIP, Video)
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
VoIP Traffic Delay with and w/o APSD
12
UL with APSD
DL with APSD
10
UL w/o APSD
Delay (millseconds)
DL w/o APSD
8
6
4
2
0
11
STA
22
STA
33
STA
4 4
STA
(VoIP only)
(VoIP only)
(VoIP, TCP)
(VoIP, Video)
Note: Jitter for VoIP traffic with APSD is < 2 µs, and <0.1 µs without APSD
Submission
Slide 11
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
Video Traffic Performance
0.9
STA4 (VoIP, Video)
STA5 (Video, TCP)
0.8
Delay (milliseconds)
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Submission
1
Uplink
2
Downlink
Uplink
3
Downlink
4
w/ APSD
w/ APSD
w/o APSD
w/o APSD
Slide 12
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
APSD’s effect on TCP throughput
1800
TCP w/ APSD
TCP w/o APSD
1600
TCP Throughput (Kbps)
1400
1200
1000
800
600
400
200
0
1
STA3
2
STA5
(VoIP, TCP)
(Video, TCP)
Note: all TCP traffic is uplink traffic
Submission
Slide 13
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
Delay Comparison for all Traffic Types
12
UL w/ APSD
DL w/ APSD
10
Delay (milliseconds)
UL w/o APSD
DL w/o APSD
8
6
4
2
0
1
VoIP
2
Video
3
TCP
Traffic Types
Note: 1) the data here are average numbers over 4 VoIP sessions, 2 Video sessions and 2 TCP
sessions. 2) there is no data collected for downlink TCP traffic so the data is not available.
Submission
Slide 14
Z. Chen, C. Zhu et al
Nov 2009
doc.: IEEE 802.11-09/1263r1
Summary and Future Works
•
Summary
– A single stream of 500Mbps cannot support the given traffic load. Aggregation gets
the job done.
– APSD allows STAs sleep >95% of the time for VoIP traffic in this specific
scenario;
– APSD increases the delay for downlink VoIP traffic (from <1ms to ~10ms), but
still within the requirement (<30ms).
– APSD slightly increases the delay for video traffic when it is running in the AP.
– TCP delay is >20% larger, and throughput is about 10% lower when APSD is
running.
•
Future work
– Run the simulation for PSMP (simulator is ready)
– Develop module to support direct links so that STA to STA traffic can be
simulated.
– Support multiple streams (receivers)
• Enhanced MAC functions needed.
Submission
Slide 15
Z. Chen, C. Zhu et al