WKH4XDOLW\RI9RLFHRYHU,3 LVLWWKHUH

Transcription

1 WKHXDOLW\RI9RLFHRYHU,3 LVLWWKHUH The Goal: Introduction to the quality issues of VoIP. Questions to be answered: What are the characteristics of the quality of VoIP? What are the loss, delay and delay variance characteristics of VoIP in an IP switching environment? How much better is the quality of VoIP over IP switching than? What is the error of delay variance estimators designed for best-effort Internet in a campus IP-switching network/ network/ethernet? 9RLFHRYHU,3 Internet GW GW PSTN/ ISDN/ GSM/ TDM PSTN/ ISDN/ GSM/ TDM

2 9RLFHRYHU,3 Coding of samples e.g. GSM. coding bit stream MEMORY BUFFER Phys. IP UDP RTP frame audio data MEMORY BUFFER bit stream Decoding of coded samples,,(& Quality of under three concepts: The characteristics of QoS Management of QoS Agreements of QoS The characteristics of QoS are quantities: throughput, prosessing delay, network end-to-end delay, delay variance, error rate, packet loss rate

3 V\QFKURQL]DWLRQ adjusting the playout time Using RTP-sender timestamps: p ki = t k i + fixed delay estimate + variable delay estimate RTP: Interarrival jitter the mean deviation of the difference in packet spacing at the receiver compared to the sender for a pair of packets Difference in packet spacing: D(k i, k j ) = (a k j - a ki ) - (t k j - t ki ) various algorithms exist for etimating Jitter from D sender VRPHWKHRU\ k-th talkspurt silence (k+)-th talkspurt t k t k t n(k) k t k+ t k+ t n (k+) k+ a k a k a k n(k) a k+ a k+ a k+ n (k+) receiver arrival p k (A) p k (A) p k n(k) (A) p k+ (A) p k+ (A) p k+ n(k+) (A) playout t k i : sender timestamp of i-th packet in k-th talkspurt a ki : receiver timestamp of i-th packet in k-th talkspurt p ki : the playout time of i-th packet in k-th talkspurt 3

4 ,3SXKHHWKHUQHWLVVl Function Delay generatot meter (7+(5(7 RADCOM LAN Source Analyzer - SUN Ultra - analysis/stats - Nevot - load Receiver - SUN Ultra - Nevot - Nevot debug - RTP dump Terminal delay: viive HW.9 ms VoIP Client ms Timestamps (sender, arrival) End-to-end delay Packet len. delay. s 5 ms. s 95 ms End End to end delay IP s&f ms IP kytkentä ms Network Network delay (ping) for both - ms 9R,3RYHU,3 BaseT IP Switch Gateway ATM 55Mbit/s ATM Switch IP Switch Gateway Source - flood ping - kilent-server - SUN Workstation - Nevot IP switch processor Delay measurement Receiver - SUN Workstation - Nevot - Nevot debug - RTP dump Receiver timestamps

5 (QGWR(QG'HOD\V 3&ZLWKHWPHHWLQJ Setup Using Microsoft Netmeeting, Windows 95 Newest DirectX version 5. (speeds up coding and decoding) Sounblaster plug&pray Pentium 33 to Pentium MHz HW delay inside PC mike to speaker: 5, ms End to End delay over LANE 7-9 ms ping delay ms,3zlwkqrordg 35 3 s&f 5 5 number of incidents IPS number of incidents 5

6 (WKHUQHWORDGZODUJHSDFNHWV Load frames/s Mbit/s ms - % 5 % % 5 % % 5 % 3 % 35 % % 5 % 5 % % % 5 % 7 % 75 % % 5 % 9 % 95 % % % Number of incidents ms - % D < n ms loaded with flood ping,5,,5,,5,3,35,,5,5,55,,5,7,75,,5,9,95 % D percentile,3vzlwfklqj ORDGHGZLWKIORRGSLQJ Number of Incidents IP switching Loaded with flood ping -, -,79 -,7 -, -,5 -, -,37 -,9 -, -, -, J (ms),,,,9,3,,5 D freq.,3,7, 5 3 Number of Incidents ping load x Mbit/s ran from edge routers large packets -,5 -,73 -,5 -,5 -,75 -,39 -,35 -, -,35 -,5,35,,5,9,375,,55,3,75, D frequency

7 LSVZLWFKLQJ7&3,3ORDG IUHTXHQFLHV Load with a perl-script: ran on edge routers sends 5 byte TCP/IP packets Load: processes 3 IP switching Load processes 7 5 Number of Incidents 5 number of incidents ,3 -,7 -, -,57 -,9 -, -,3 -,3 -, -,,,,9,,37,5,5,,7, Frequency -,5 -,73 -,5 -,5 -,75 -,39 -,35 -, -,35 -,5,35,,5,9,375,,55,3,75, Frequency 5 LSVZLWFKLQJ7&3,3ORDG QHWZRUNORDGLQPHDVXUHPHQW Traffic during measurement 9 Traffic during measurement IP Switching 7 Net kbits/s 5 User kbit/s Net (Kbits) User (Kbits) :: 3::33 3::3 3::3 3:: 3::53 3::5 3:3:3 3:3: 3:3:3 3:3: 3:3:3 3:3: 3:3:33 3:3:3 3:3:3 3:3: 3:3:53 3:3:5 3::3 3:: 3::3 3:7: 3:7: 3:7:3 3:7: 3:7:5 3:: 3:: 3:: 3:: 3::3 3:: 3:: 3::5 3:9: 3:9: 3:9: Mean Minimum Maximum 79 Kbit/s 79 Kbit/s 777 Kbit/s Mean Minimum Maximum Kbit/s 7 Kbit/s 7 Kbit/s 7

8 ,3SURFHVVHV IP switching D EJ_ - D EJ_ - packets - packets LSVZLWFKLQJ 7&3,3ORDGVXPPDU\ Differences in packet spacings of voice stream: IP Forwarding Standard Deviation, Sample Variance 7,9 Minimum -9,77 Maximum 5,5 IP Switching Standard Deviation, Sample Variance,9 Minimum -9,3 Maximum,3

9 LSVZLWFKLQJ7&3,3ORDG ªSDFNHWVSDFLQJV Load with a client-server application: perl scripts run on edge routers sends 5 byte TCP/IP packets x sending processes Jitter of voice stream, processes IP switching Jitter of voice stream IP switching, processes jitter (ms) packets J EJ_ jitter (ms) packets J EJ_ IP switching Load processes LSVZLWFKLQJ7&3,3ORDG ªIUHTXHQFLHV 3 processes number of incidents 5 5 number of incidents -,5 -,775 -,7 -,5 -,55 -,75 -, -,35 -,5 -,75 -, -,5 jitter (ms),5,5,,75,35,5,5,575,5,75, Frequency -,3 -,7 -, -,57 -,9 -, -,3 -,3 -, -,,, jitter (ms),9,,37,5,5,,7, Frequency IP switching Standard Deviation, Sample Variance, Minimum -9,77 Maximum,3 Standard Deviation,3 Sample Variance,9 Minimum -9,7 Maximum,3 9

10 ,3PL[HGORDG Number of Incidents -, -, -,7 -,7 -, -,5 -,5 -, -, -,3 ms -, -, -, -, -,,,,,,,3,3,,5,5,,,7, - % 5 % % 5 % % 5 % 3 % 35 % % 5 % 5 % 55 % % % 5 % 7 % 75 % % 5 % 9 % 95 % % % D < n ms,3zlwfklqjpl[hgordg JHWWLQµWKDWODVWE\WH Differences in packet spacings of voice stream: Standard Deviation,93 Sample Variance,5 Minimum -9,77 Maximum,7 Jitter of voice stream IP switching, mixed load IP switching mixed load jitter (ms) packets J EJ_

11 ,3SXKH,3N\WNHW\VVlYHUNRVVD D(ms),,, 5, % 59, %, % 5, %, % 7, % 7, % 77, %, % % 3, %, % 9, % 9, % 95, % 9, % (VWLPDDWWLHQYLUKHHW 5 % % % % % % % % 3 % 3 % % % % 5 % 5 % % % % 7 % 7 % % % % 9 % Error of estimate 9 % % C=/ C=/ C=/ C=/ -5 %'

12 &RQFOXVLRQV End-to-end delays caused by poor terminal performance => functions on hw At its best delays acceptable, around ms (GSM ms) IP switching reduces network caused jitter of audio stream when most traffic switched proritizing increases differences no differences in fixed delay (network too small) IP switching increases throughput of network compared to but not to wire-speeds yet Even with worst loads all measured networks good for VoIP Even the generic interarrival jitter estimation algorithm ok, although the weighting factor in RTP not optimal => adaptive algorithms with error feedback 9RLFHRYHU,3HDVXUHPHQWV Contents: Concepts: delay, delay variance (jitter), synchronization Measurement setup End-to-end delay measurements Delay variance of a voice stream over Ethernet Delay variance of a voice stream over IP switching Conclusions