Transcript Voice over IP (Lecture 15)

Voice over IP
• Why
• Challenges/solutions
• Voice codec and packet delay
• Motivation:
– Benefits:
• Reduce backbone network costs: managing a single
packet backbone instead of multiple backbones
(packet switching for IP and circuit switching for
voice).
– No way for TDM networks to support IP traffic
• Reduce access network costs:
– Bandwidth saving
– one access line for all services
• Reduce premise network (local area network) costs:
– Use one network to do everything.
• Challenges:
– Bandwidth management to support carrier
grade phone calls – really need working IP QoS
mechanism.
– Signaling
• Functionality in telephone system is now very
complicated. Everything must be re-engineered in
the corresponding signaling system in IP network.
SIP and H.323
– Media transport
• Need a protocol to transport the contents. Real Time
Protocol (RTP).
– Interoperability: work with the POTS.
• VoIP and QoS:
– Major challenges: delay and delay variation(Jitter).
– Voice applications are usually interactive.
• delay requirement for a telephone system: 150ms-250ms.
– The sources of delay in a voice over IP system:
• OS delay: 10s-100s milliseconds
• Voice processing delay: DSP 10s milliseconds, Sound cards:
20-100 milliseconds.
• Look-ahead processing delay: coding may need to know the
next few samples (5ms-7.5ms).
• Packetization delay for voice samples: multiple sample are
usually packed into a packet to save bandwidth.
– (n-1)*0.125us: 40 * 0.125 = 50ms
• Packetization delay for voice packet: (n-1)t, can be quite large.
• Modem delay: 20-40ms per modem.
• The sources of delay in a voice over IP system
(continue):
• Ingress/egress delay: transmission delay at the access line. 50
bytes on a 33Kbps access line: 50 * 8 / 33 = 12 ms
• Network delay: 15ms propagation delay for 3000km wires.
100ms all together.
– Total delay:
• Gateway to gateway: roughly 180ms (100ms network delay).
• Desktop to desktop: roughly 450ms.
– Delay control mechanism: network priority
mechanisms, end hosts priority mechanism, edge
equipment design (IP QoS + Real time Operating
Systems + voice hardware)
• Source jitter:
– Network: network conditions vary at different times.
– Non-real time OS: samples processed at different time.
• Jitter control: buffering at the destination.
• QoS parameters:
–
–
–
–
Accuracy
Latency
Jitter
Codec quality
• QoS control mechanisms: sender-based, networkbased and receiver-base
• Sender-based:
– Retransmissions
– Forward error correction
– Interleaving
• Receiver-based:
– Switching to lower bandwidth encoding
– Concealment (silence insertion, noise insertion, repeat
previous packet, repeat and fade, interpolate).
• Network-based: IP QoS
• Voice codes/packet delay and RSVP:
Codec
G.711
G.722
G.726
G.726
kbps sample size(bits)
64
8
64
8
16(24…)
2(3/4/5)
16
2
no. of samples no. of bytes delay
80
80
10ms
160
160
20ms
80
20
10ms
240
60
30ms
– Issues in Media transfer:
• RTP/UDP/IP/link layer protocol
• Protocol overheads: 12 bytes RTP header, 8 bytes UDP header,
20 bytes IP header.
• G.726 16kbps encoding: 20 bytes payload. 33% link
efficiency.
• Mapping voice stream into TSpec in RSVP
G.726 16kbps encoding with a packet time of 10 ms
TSpec: Bucket depth, b
Bucket rate: r
Peak rate: p
minimum policed unit: m
Maximum packet size: M
How to map?
• Reducing header overheads:
– Frame packing:
• More frames in one packet
– Less overhead
– Less number of total packets in the system
• Problem?
– RTP multiplexing:
• Put multiple frames from different calls in one packet
– RTP header compression
• Most fields in the headers are fixed throughout a session.
• Record a context id in each router and use the id to decide what
to do. Reduce RTP/UDP/IP headers to 10 bytes.
• Need path setup
• No longer native IP packets.