Affects of Queuing Mechanisms on RTP Traffic Comparative Analysis of - - PowerPoint PPT Presentation

Aim and Objectives Introduction Simulations and Results Conclusion References

Affects of Queuing Mechanisms on RTP Traffic Comparative Analysis of Jitter, End-to- End Delay and Packet Loss

Gregory Epiphaniou1 Carsten Maple1 Paul Sant1 Matthew Reeves2

1Institute for Research in Applicable Computing

University of Bedfordshire

2Modern Networks

Hitchin, United Kingdom

ARES Conference, 2010

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References

Outline

1

Aim and Objectives

2

Introduction Some facts on VoIP implementation VoIP Impairments Queuing Mechanims

3

Simulations and Results First-In-First Out Random Early Detection Differentiated Services

4

Conclusion

5

References

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References

Aim and Objectives

Investigate VoIP traffic behaviour under burst traffic conditions Illustrate precise effects of queuing mechanisms to VoIP and compare results Find the most appropriate mechanism to be used in the case of unelastic traffic DropTail (FIFO), RED and DiffServ, and their effects on real-time voice traffic Extract measurements on e2e delay, jitter and packet loss rates for each mechanism examined The NS-2 simulator has been used

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References Some facts on VoIP implementation VoIP Impairments Queuing Mechanims

Some facts on VoIP implementation

IP telephony (IPTel) refers to the technology to transport real-time media

• ver an IP network

Converging voice and data into a best-effort service network demands preferential handling of traffic QoS assurance has increased due to the enormous growth of users accessing networks Voice coders contribute to end-to-end delay due to sample accumulation delays and look-ahead delays VoIP must reach the QoS that a Public Switched Telephone Network (PSTN) provides Best-effort service networks has resulted in many technical challenges regarding traffic engineering and shaping

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References Some facts on VoIP implementation VoIP Impairments Queuing Mechanims

VoIP Impairements – Jitter and E2E Delay

Jitter can be effectively described as the unwanted variation in the inter-arrival packet time Jitter = (Ti − Ti−1) − (Tsi − Tsi−1) (1)

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References Some facts on VoIP implementation VoIP Impairments Queuing Mechanims

VoIP Impairements – Jitter and E2E Delay

Jitter can be effectively described as the unwanted variation in the inter-arrival packet time Jitter = (Ti − Ti−1) − (Tsi − Tsi−1) (1) End-to-end delay is one of the most important threats to perceived QoS Dk =

H

• i=1

= L Ci − qk

i

Ci

• =

H

• i=1

L Ci + dk

i

(2)

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References Some facts on VoIP implementation VoIP Impairments Queuing Mechanims

VoIP Impairements – Packet Loss Rate and NIST Recommendations

Packet loss may occur at any stage of a network transmission Plrij = pktslost pktsreceived = pktssenti − pktsreceivedj pktsreceivedj (3) PLRij ≤ 0.05 ∗ Pktssenti (4) The ITU-T (International Telecommunication Union) recommends that for

• ne way transmission the actual end-to-end delay limits are:

0 to 150ms: Acceptable for most network hosted applications. 150 to 400ms: International or satellite connections. > 400ms: Unacceptable for general network purposes.

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References Some facts on VoIP implementation VoIP Impairments Queuing Mechanims

VoIP Impairements – Media Path

Additional Delays Packetisation Delay TDP = Ps Cbw (5) Serialisation Delay Ds = Ps + HL Ls (6) Propagation Delay Dpr = L u (7)

Figure: VoIP media path [?]

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References Some facts on VoIP implementation VoIP Impairments Queuing Mechanims

Queuing Mechanisms - FIFO

In FIFO all packets are treated equally and served at the same

• rder in which they were placed in

the queue

Figure: FIFO Queue

Computationaly inexpensive Predictable queue behaviour since no packet re-ordering takes place Minimum queuing delay at the intermediate hops Does not support traffic clasification Linear relationship between aggregation of incoming traffic and mean queuing delay Benefits UDP flows

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References Some facts on VoIP implementation VoIP Impairments Queuing Mechanims

Queuing Mechanisms - RED

Figure: RED Queue

Counts the dropping probability of a packet Decisions are made based on Max/Min queue length threshold If the avg queue length is greater than max threshold the packets are marked The average queue size does not exceed the maximum threshold High throughput and low avg delay for high speed networks with TCP connections

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References Some facts on VoIP implementation VoIP Impairments Queuing Mechanims

DiffServ

Figure: IPv4 ToS byte Figure: DiffServ code point field

Achieves division of packets into classes by using a 6-bit DSCP Define the policies for the level of service for each flow Performs packet marking and traffic shaping based on the policies Core routers forward packets based on their marking

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References First-In-First Out Random Early Detection Differentiated Services

Simulation Parameters

Elastic and unelastic traffic examined (RTP and TCP) Pareto ON/OFF model has been used to mimick VoIP traffic G.711 (64kbps), burst time 20ms, idle time 10ms and rate 87.4kbps Default payload size of 160 bytes Total simulation time 62sec No cRTP/Multiplexing mechanisms

Table: Link characteristics

nodes Propagation Delay [ms] Link capacity [Mb] n0 - n2 10 1 n1 - n2 10 1.5 n2 - n3 20 2 n3 - n4 10 1 n3 - n5 10 1

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References First-In-First Out Random Early Detection Differentiated Services

FIFO Results and Discussion

Figure: Simulation Model FIFO

Traffic sent Dropped TCP 10484 5 RTP 101771 1262

10 20 30 40 50 60 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 Simulation time [sec] Delay variation for RTP traffic with FIFO [sec]

Figure: Jitter for RTP traffic with FIFO

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References First-In-First Out Random Early Detection Differentiated Services

FIFO e2e delay

10 20 30 40 50 60 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2 Simulation time [sec] RTP OWD for RTP traffic with 50pkts buffer size [sec]

Figure: Simulation Model FIFO: RTP One Way Delay with buffer size of 50pkts and DropTail

10 20 30 40 50 60 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 Simulation time [sec] RTP OWD for RTP traffic with 25pkts buffer size [sec]

Figure: Simulation Model FIFO: RTP One Way Delay with buffer size of 25pkts and DropTail

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References First-In-First Out Random Early Detection Differentiated Services

RED Results and Discussion

Figure: Simulation Model RED

10 20 30 40 50 60 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 Simulation time [sec] Delay Variation for RTP traffic with RED [sec]

Figure: Simulation Model RED: Delay Variation for RTP traffic against simulation time

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References First-In-First Out Random Early Detection Differentiated Services

RED e2e delay and packet loss

10 20 30 40 50 60 100 200 300 400 500 600 Simulation time [sec] Total RTP packets dropped with RED

Figure: Simulation Model RED: RTP packet drop rate against simulation time

10 20 30 40 50 60 0.05 0.1 0.15 0.2 0.25 Simulation time [sec] RTP OWD with RED [sec]

Figure: Simulation Model RED: RTP One Way Delay with buffer size of 50pkts against simulation time

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References First-In-First Out Random Early Detection Differentiated Services

DiffServ Results and Discussion

Figure: Simulation Model DiffServ

Traffic Sent Dropped TCP 9184 4 RTP 134564 864

10 20 30 40 50 60 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 Simulation time [sec] Delay variation for RTP traffic with DiffServ

Figure: Jitter for RTP traffic against simulation time

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References First-In-First Out Random Early Detection Differentiated Services

DiffServ e2e delay and packet loss

10 20 30 40 50 60 100 200 300 400 500 600 700 800 900 Simulation time [sec] Total RTP packets dropped with DiffServ

Figure: Simulation Model DiffServ: RTP packet drop rate against simulation time

10 20 30 40 50 60 0.02 0.04 0.06 0.08 0.1 0.12 0.14 Simulation time [sec] RTP OWD with DiffServ

Figure: Simulation Model DiffServ: RTP One Way Delay with buffer size of 50pkts against simulation time

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References

Conclusion

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic

Aim and Objectives Introduction Simulations and Results Conclusion References

References

Epiphaniou,Maple,Sant,Reeves Affects of Queuing Mechanisms on RTP Traffic