Multimedia Communications Spring 2006-07 Delay of Voice Traffic - PowerPoint PPT Presentation

CS 584 / CMPE 584 Multimedia Communications Spring 2006-07 Delay of Voice Traffic Over IP Netw orks Shahab Baqai LUMS

Characteristics and Requirements of Voice Traffic � Voice characteristics – Low rates (8Kb/s for G.729A, 64Kb/s for G.711) – Low variability resulting from Silence Suppression Talk Spurt Silence λ μ � Voice requirements: real-time communication – Low delay: 200-300ms round-trip, 100-150ms one-way (Dmax) – Low jitter: for smooth playback – Low packet loss: at most 2% 2

Tolerable Packet Loss Versus Packet Formation Time -1 10 Maximum Tolerable Packet Loss Independent packet loss -2 10 -3 10 -4 10 -5 10 Correlated packet loss -6 10 0 1 2 10 10 10 Packet Formation Time T f (ms) 3

Case for Separation of Voice From Other Traffic � Measurements on the Internet as well as simulations show that mixing voice and TCP data traffic leads to either – Large delays for voice (100-500ms average delay) – Very low utilization of network resources (often less than 20%) � Voice must also be separated from bursty UDP traffic (e.g., VBR video) – Voice delay affected if aggregate peak rate exceeds available bandwidth – Results in low bandwidth utilization 4

Case for Separation of Voice From Other Traffic 10Mb/s, 7hops 1 12 � VBR video stream (730 Kb/s 10 -1 average, 2.3Mb/s peak) 12 � 450 Kb/s aggregate voice 7 load 10 -2 Delay CCDF VBR Video Voice mixed with 10 -3 video 10 -4 Voice separated from 1-4 video 10 streams video 10 -5 0.1 1 10 100 Delay (ms) 5

Analysis of Voice Delay and Jitter � Voice better served if separated from other traffic in the network – Voice given its own circuit – Voice given priority over other traffic – Voice given its share of the bandwidth using WRR � In this context, study effect of – Residual transmission time of lower priority traffic – Available bandwidth – Scheduling scheme – Packet formation time 6

End-to-end Delay Components Sender Receiver Packetization De-packetization Encoder Decoder Network Playback Buffer • Delay budget ∑ = + + + + + ≤ ( ) D T l T P Q D D f i i i play max ∈ i path 7

Encoding and Packetization Analog Signal Frame f Encoder In Lookahead l Processing p e Encoder Out Encoding Delay D encoder Packetization Out Encoding and Packetization Delay D encoder +D pack 8

End-to-end Delay Components ∑ = + + + − + + + + + D f l p ( k 1 ) f ( Q T P ) D p e i i i play d i ∑ = + + + + + + + kf l p ( Q T P ) D p e i i i play d i ≤ , (can be assumed negligible ) p p f e d ∑ = + + + + + ( ) D kf l Q T P D i i i play i T f = Packet formation time D net = Network delay 9

End-to-end Delay Components ∑ ∑ = + + + + + ( ) D T l T P Q D f i i i play ∈ ∈ i path i path Source of Jitter, random, function of: Constant, Constant, • Path in the network function of: function of: • Traffic load and • Path in the network •Encoding Scheme characteristics (links speed, links (frame size, look- • Scheduling scheme propagation delay) ahead) • Packet size • Packet size •Packet size 10

Playback Buffer � Jitter unknown and random � Thus, playback buffer needed – Packets delayed by playback buffer delay D play – Insures continuous playback � D play depends whether sender and receiver clocks are – Synchronized (e.g., GPS) – Not synchronized 11

Synchronized Clocks Sender Receiver Encoding and De-packetization Network packetization and decoding t sender t receiver time t receiver -t sender D play D max − + ≤ We must have ( t t ) D D receiver sender play max 12

Synchronized Clocks, Unknown Jitter � To accommodate highest amount of jitter (and minimize loss), ∑ ∑ ∑ = − − = − + + + − choose D D ( t t ) D ( T l T P ) Q play max rec send max f i i i i i i ∑ ∑ ∑ = = = − + + + max If Q 0 (no jitter) then D D D ( T l T P ) i play play max f i i i i i ∑ ∑ ≤ − max max If Q D , packet received and buffered for D Q i play play i i i ∑ Q > max If D , packet rejected i play i max D play 13

Synchronized Clock, Known Jitter � If maximum bound on jitter known (by some measure), i.e. ∑ i = max( Q ) Q max i D play = max then Q max ∑ = + + + + and D T l ( T P ) Q max f i i i max D play 14

Non-Synchronized Clocks � RTP timestamps packets � However, no guarantee that sender and receiver are synchronized � Hence, receiver does not know at what time the packet was sent 15

Play-out Buffer Delay When first packet received at the destination, it must be delayed by Q max in order to take into account the jitter Σ i Q i ∑ = = max ( ) D Q Q max play path i ∈ i path ... Sender ... Receiver Q max ... Display D play = Q max 16

End-To-End Delay Requirements • Hence, the one-way end-to-end delay becomes − ⎧ 2 Q , S and D non synchroniz ed ∑ = + + + + max ⎨ D T l ( T P ) f i i ⎩ Q , S and D synchroniz ed ∈ i path max � Toll quality real-time communication needed – Round-trip delay must be in the range 200-300 ms – That is, D ≤ D max , where 100 ms ≤ D max ≤ 150 ms � Amount of jitter allowed 10-50 ms, function of – Acceptable end-to-end delay D max – Formation time T f – Propagation delay 17

Network Scenario � Assess queuing delay incurred Target Source by a voice stream traveling through a given number of hops � Consider hops to be independent � Delay observed depends on characteristics of interfering traffic Target Receiver 18

Sources of Jitter � Queuing delay behind voice packets in the same queue – Depends on the burstiness of the voice traffic pattern (Variability of traffic at the source introduced by Silence Suppression) � Residual transmission time of lower priority packets – Increases the burstiness of the outgoing voice streams 19

Models for Voice Traffic � Voice traffic alone on the network – Voice traffic does not incur transmission time from lower priority packets – Traffic variability minimal – One hop: queuing delay Σ D i /D/1 � Even though Silence Suppression reduces voice rate, effect on delay negligible – H hops: convolution of one hop delays 20

Models for Voice Traffic (cont.) � Voice traffic incurs residual transmission time of lower priority packets – Large traffic variability possible � Decreases with the number of hops � Tail of inter-arrival time between packets bounded by exponential distribution – One hop: delay percentile is sum of � Queuing delay percentile M/D/1 � Delay percentile from (uniformly distributed) residual transmission time – H hops: delay percentile is sum of � Delay percentile from convolution of M/D/1 queuing delays � Delay percentile from convolution of transmission times 21

Models Versus Simulation Results Q CCDF, 4 streams, 5 hops, 384Kb/s Modeling 1 Simulation 10 -1 Voice Alone 10 -2 Residual 10 -3 transmission time of lower priority packets incurred 10 -4 1 10 100 Queuing Delay Q (ms) 22

Effect of Residual Transmission Time of Lower Priority Packets CCDF of Queuing Delay T1, G.729A, T f =30ms, 50% utilization 1 1 hop 2 hops 5 hops 10 -2 Voice Alone 10 -4 Residual transmission time of lower priority packets incurred 10 -6 0.01 0.1 1 10 100 Queuing Delay (ms) 23

Effect of Bandwidth CCDF of Queuing Delay G.729A, T f =30ms, 50% utilization 1 10 -2 T1 T3 10 -4 5 hops 1 hop 10 hops 1 hop 10 -6 0.1 1 10 Queuing Delay (ms) 24

Choice of Scheduling Scheme CCDF of Queuing Delay Weighted Round Robin (WRR) Versus Priority Queuing (PQ) 1 WRR, 1.5Mb/s 10 -2 WRR, PQ 10Mb/s 10 -4 T3, G.729A, T f =30ms, 5 hops, 1.35Mb/s Voice Load 10 -6 0.1 1 10 100 Queuing Delay (ms) 25

Choice of Packet Formation Time Data Link/MAC IP 20 Bytes UDP 8 Bytes RTP 12 Bytes Encoded Voice 4 Total Header H=46-69Bytes 3.5 3 2.5 � r = rate of encoded bit R/r 2 � stream 1.5 G.729A � 1 R = rate of packetized bitstream G.711 0.5 0 8 H = + 0 20 40 60 80 100 R r Formation Time T f T f � Incentive to use largest formation time possible (given D max , propagation and queuing delays) 26

Effective Header Size, IEEE 802.11 Wireless Network Physical Layer No RUI H. Comp. RUI H. Comp. DSSS, 1Mb/s 97 Bytes 67 Bytes DSSS, 2Mb/s 120 Bytes 90 Bytes 802.11b, 5.5Mb/s 200.5 Bytes 170.5 Bytes 802.11b, 11Mb/s 327 Bytes 297 Bytes 802.11a, 6Mb/s 89 Bytes 59 Bytes 802.11a, 24Mb/s 134 Bytes 104 Bytes 802.11a, 54Mb/s 209 Bytes 174 Bytes � Physical layer header must be transmitted at lowest speed – Hence, effective header size increases with available bandwidth 27

100 80 Formation Time T f 60 802.11 T1/T3 802.11b 40 20 28 Packetization Overhead 0 20 15 10 5 0 R/r G.723.1 G.729A G.711