comparison and analysis of fifo, pq, and wfq disciplines ... › ~ljilja › ensc427 › spring11...
TRANSCRIPT
ENSC 427
Communication Networks Final Project Demo – Spring 2011
1
Comparison and analysis of FIFO, PQ, and WFQ Disciplines in OPNET
2
Introduction and Motivation • Queuing methods are used to handle network resources
• Real-time applications such as voice and video conferencing are
especially susceptible to delays and packet losses
• The Quality of Service (QoS) network devices must employ proper queuing methods to differentiate among arriving packets
• Queuing disciplines are implemented in routers
• We seek to provide an answer to the question:
• What is the best queuing discipline to use for a given application?
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
Comparison and analysis of queuing disciplines in OPNET
• Focus on:
▫ FIFO First-in, First Out
▫ PQ Priority Queuing
▫ WFQ Weighted Fair Queuing
• Apply each queuing method towards:
▫ FTP
▫ Voice
▫ Video Conferencing
3
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
Overview
4
Global and Object statistics collected and analyzed
Overview
Statistics for Object Statistics
Point-to-point
Average Queuing Delay ->
Throughput ->
Utilization ->
Statistics for Global Statistics
FTP
Traffic Sent
Traffic Received
VoIP
End-to-End Delay
Jitter
Mean Opinion Score
Packet Delay Variation
Traffic Sent
Traffic Received
Video Conferencing
End-to-End Delay
Packet Delay Variation
Traffic Sent
Traffic Received
IP Traffic Dropped
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
5
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
OPNET Implementation
Scenario 1
Scenario 2
Scenario 3
Queuing discipline
First-in, First-out Queuing
Priority Queuing
Weighted Fair Queuing
Simulation Time 5 minutes 5 minutes 5 minutes
Applications
considered
FTP
VoIP
Video Conferencing
FTP
VoIP
Video Conferencing
FTP
VoIP
Video Conferencing
Scenarios and applications considered
6
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
OPNET Implementation • Network Topology
• Campus network 10 x 10 km
• 5 workstations
• 2 routers
• 1 Ethernet server
• Application, Profile, and QoS
Definitions
7
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
Application Definition Attributes
Application Name
FTP
VoIP
Video Conferencing
Description
High Load
PCM Quality Speech
Low Resolution Video
Type of Service (ToS)
Best Effort (0)
Interactive Voice (6)
Streaming Multimedia (4)
OPNET Implementation
Application Attributes
8
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
FTP
Simulation Results –Traffic Sent
• Traffic sent in bytes/s
• Application : FTP
• FIFO and WFQ show reasonable traffic sent
• PQ worst queuing discipline for FTP traffic with respect to traffic sent
9
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
FTP
Simulation Results –Traffic Received • Traffic received in bytes/s
• Application : FTP
• WFQ shows more traffic
received than FIFO over the long run
• PQ worst queuing discipline for FTP traffic, no traffic is received
• FTP Packets given lowest priority
• FTP traffic not sensitive to jitter, delay
Voice
Simulation Results – End-to-End Delay
10
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
• Time taken for packets to be transmitted from source to destination.
• Application : Voice
• ETE in voice and video should be small to provide natural conversation
• FIFO shows most ETE Delay (~ 2 sec)
• PQ and WFQ have lower ETE delay(~ 0.063 sec)
11
Voice
Simulation Results - Jitter
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
• Jitter is the variation in ETE Delay
• Application : Voice
• Jitter should be minimized especially in real time applications
• FIFO shows most jitter
• PQ and WFQ show less jitter than FIFO
12
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
Voice
Simulation Results – MOS Value • Mean Opinion Score
defines the perceived voice quality
• MOS scale : 1-5 • 1: Bad • 2: Poor • 3: Fair • 4: Good • 5: Excellent
• FIFO shows bad perceived
audio quality
• PQ and WFQ in between fair and good perceived audio quality
13
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
• PDV is a measure of the difference in ETE
• Ignores lost packets
• Application : Voice
• PDV should be minimized especially in real time applications
• FIFO shows most PDV
• PQ and WFQ show less PDV than FIFO
Voice
Simulation Results – Packet Delay Variation
14
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
Voice
Simulation Results – Traffic Received
• Voice Traffic received in bytes/s
• Traffic received roughly equal initially
• Degree of loss increases over time
• FIFO queuing results in fewer bytes received
• PQ and WFQ result in more traffic received
15
Simulation Results – IP Traffic Dropped
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
• IP traffic dropped packets/sec
• Drop in IP Traffic results due to insufficient queue space
• PQ and WFQ shows less drop in IP packets
• FIFO shows most drop
• Drop should be minimized, especially in real-time applications
16
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
Video Conferencing
Simulation Results – End-to-End Delay • ETE delay (s)
• WFQ shows highest
ETE because it transmits packets from all applications fairly
• FIFO shows no preference for packets sent, shows better ETE Delay than WFQ
• PQ shows smallest ETE, it is the best queuing choice for video conferencing with respect to ETE Delay
17
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
Video Conferencing
Simulation Results – Packet Delay Variation • PDV is a measure of the
difference in ETE, it ignores lost packets
• Application : Video
• PDV should be minimized especially in real time applications
• WFQ shows most PDV
• FIFO better choice than WFQ
• PQ is best choice for video with respect to PDV
18
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
Video Conferencing
Simulation Results – Traffic Received • Video Traffic received in
bytes/s
• PQ results show it is worst queuing technique for video with respect to traffic received
• FIFO better choice than PQ but has only 1 queue and no priority given to video packets
• WFQ chosen as best queuing discipline because it prioritizes packets
19
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
Point-to-Point: Router A -> Router B
Simulation Results – Average Queuing Delay
• Average Queuing Delay in (s)
• FIFO shows worst Average Queuing Delay
• PQ results show it is much better as compared with FIFO
• WFQ chosen as best queuing discipline in regard to the Average Queuing Delay
20
• Throughput (packets/s) is the average number of packets received and transmitted by receiver and transmitter channels per second
• FIFO shows worst Throughput
• PQ results show it is much better as compared with FIFO
• WFQ chosen as best queuing discipline with regard to Throughput
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
Point-to-Point: Router A -> Router B
Simulation Results – Throughput
21
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
Comparison of Results Obtained
Ranking
System
Equivalent
Numerical Value
Good 1
Worse 2
Worst 3
Ranking System Employed
Statistics
Collected
Scenario
1: FIFO
Scenario
2: PQ
Scenario
3: WFQ
Best Queuing
Discipline
IP Packets
Dropped
3 2 1 WFQ
Average Queuing
Delay
2 1 1 WFQ or PQ
Utilization 1 1 1 FIFO, WFQ, or
PQ
Throughput 3 2 1 WFQ
Comparison of Collected Statistics, part 1
22
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
Comparison of Results Obtained
Application Statistics
Collected
Scenario
1:
FIFO
Scenario
2: PQ
Scenario 3:
WFQ
Best Queuing
Discipline
FTP Traffic Sent 2 3 1 WFQ
Traffic Received 2 3 1 WFQ
Voice
End-to-End Delay 2 1 1 WFQ or PQ
Jitter 3 1 2 PQ
Mean Opinion Score 3 1 2 PQ
Packet Delay
Variation
2 1 1 WFQ or PQ
Traffic Sent 1 1 1 FIFO, WFQ, or
PQ
Traffic Received 3 1 2 PQ
Video
Conferencing
End-to-End Delay 2 1 3 PQ
Packet Delay
Variation
2 1 3 PQ
Traffic Sent 1 1 1 FIFO, WFQ, or
PQ
Traffic Received 2 3 1 WFQ
Comparison of Collected Statistics, part 2
23
Conclusion
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
• Voice Applications: • Best: Priority Queues • Worst: FIFO
• Video Applications:
• Best: Weighted Fair Queues • Worst: Priority Queues
• FTP Applications:
• Best: Weighted Fair Queues • Worst: Priority Queues
24
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
Future Work
• Should study effects of other Queuing Disciplines such as DWRR, Custom Queues, SPQ, and SFQ
• Should study effects of Random-Early Drop (RED) and Drop-tail Policy
• Should consider various other applications such as online gaming.
• Should consider different voice and video qualities to better understand and justify best type of Queue to choose
25
References
ENSC 427 – Communication Networks – Final Project Professor: Ljiljana Trajković
[1] S.P. Morgan. "Queueing Disciplines and Passive Congestion Control in Byte-Stream Networks." in INFOCOM '89 Proc. 8th Annu. Joint Conference IEEE Computer and Communications Societies, 1989, pp. 1097-1106. [2] L.G. Widjaja and I. Widjaja. Communication Networks. New York, NY: McGraw-Hill, 2004, pp. 539- 547. [3] K. James, and R. Keith. "Scheduling and Policing Mechanisms.” http://www3.gdin.edu.cn/jpkc/dzxnw/jsjkj/chapter6/6-6.htm [Apr. 3, 2011]. [4] B. Dekeris, T. Adomkus, and A. Budnikas. "Analysis of QoS Assurance using Weighted Fair Queueing (WQF) Scheduling Discipline with Low Latency Queue (LLQ)." in 28th Int. Conf. Information Technology Interfaces, 2006, pp. 507-512. [5] P. Calyam et al. "Impact of Router Queuing Disciplines on Multimedia QoE in IPTV Deployments." in QoMEx 2009 Int. Workshop Quality of Multimedia Experience, 2009, pp. 92-97. [6] S. Minseok, C. Naehyuck, and S. Heonshik. "A New Queue Discipline for Various Delay and Jitter Requirements in Real-Time Packet- Switched Networks." in Proc. 7th Int. Conf. Real-Time Computing Systems and Applications, 2000, pp. 191-198. [7] Z. Ni, X. Lu, and D. Liu. "Simulation of Queuing Systems with Different Queuing Disciplines Based on Anylogic." in Int. Conf. Electronic Commerce and Business Intelligence, 2009, pp.164-167. [8] T. Velmurugan, H. Chandra, and S. Balaji. "Comparison of Queuing Disciplines for Differentiated Services Using OPNET." In ARTCom ’09 Int. Conf. Advances in Recent Technologies in Communication and Computing, 2009, pp.744-746.