Conclusion
With regards to our criteria of low packet loss to maintain a high voice quality, we see that TCP out- performs UDP/RTP. Packet losses were not observed for TCP unless the background traffic was at
maximum load. Afterwards, only about 20 packets were dropped out of a total of 310 sent which results
in a 6.5% loss throughout the simulation. The percentage lost for UDP/RTP throughout the simulation is
much higher at approximately 25% for traffic sent from Node 0 to Node 4, however this is mostly due to
massive losses caused by the exceedingly large background traffic later into the simulation. Up until
about 30 seconds (low to moderate load), we observe a very acceptable 5% loss for traffic flowing either
way.
Nonetheless, low delay and jitter are a higher priority than loss of quality. TCP’s sluggish delay and large
jitter is unacceptable for VoIP applications and any benefits obtained in terms of improved voice quality are nullified by this problem. Conversely, the simple, “best effort” nature of UDP/RTP allows for very
small delay and jitter. This, along with acceptable voice quality during low/moderate background traffic,
is what makes UDP or RTP the protocols of choice for VoIP applications. In general, RTP is almost always
used over UDP due to the identical performance and added features.
Currently, Session Initiated Protocol (SIP) is growing in popularity as the choice signaling protocol of VoIP
networks. For future work, an implementation of SIP in NS-2 can be done and the numerous capabilities
provided by it, such as conference calls and call-switching can be simulated.
ConclusionWith regards to our criteria of low packet loss to maintain a high voice quality, we see that TCP out- performs UDP/RTP. Packet losses were not observed for TCP unless the background traffic was atmaximum load. Afterwards, only about 20 packets were dropped out of a total of 310 sent which resultsin a 6.5% loss throughout the simulation. The percentage lost for UDP/RTP throughout the simulation ismuch higher at approximately 25% for traffic sent from Node 0 to Node 4, however this is mostly due tomassive losses caused by the exceedingly large background traffic later into the simulation. Up untilabout 30 seconds (low to moderate load), we observe a very acceptable 5% loss for traffic flowing eitherway.Nonetheless, low delay and jitter are a higher priority than loss of quality. TCP’s sluggish delay and largejitter is unacceptable for VoIP applications and any benefits obtained in terms of improved voice quality are nullified by this problem. Conversely, the simple, “best effort” nature of UDP/RTP allows for verysmall delay and jitter. This, along with acceptable voice quality during low/moderate background traffic,is what makes UDP or RTP the protocols of choice for VoIP applications. In general, RTP is almost alwaysused over UDP due to the identical performance and added features.Currently, Session Initiated Protocol (SIP) is growing in popularity as the choice signaling protocol of VoIPnetworks. For future work, an implementation of SIP in NS-2 can be done and the numerous capabilitiesprovided by it, such as conference calls and call-switching can be simulated.
การแปล กรุณารอสักครู่..
