Bandwidth Engineering

To successfully implement CAC mechanisms in your packet network, you must begin with a clear understanding of the bandwidth required by each possible call that can be placed. In Chapter 7, "Link-Efficiency Tools," you learned about bandwidth requirements for two of the most popular codecs deployed in converged networks, G.711 and G.729.

The G.711 codec specification carries an uncompressed 64-kbps payload stream, known in the traditional telephony world as pulse code modulation (PCM). G.711 offers toll-quality voice conversations at the cost of bandwidth consumption. The G.711 codec is ideally suited for the situation in which bandwidth is abundant and call quality is the primary driver, such as in LAN environments.

Figure 8-3 IP Telephony Network with CAC

IP Network Supports 2 Calls Max!

Figure 8-3 IP Telephony Network with CAC

IP Network Supports 2 Calls Max!

The G.729 codec specification carries a compressed 8-kbps payload stream, known in the traditional telephony world as conjugate-structure algebraic-code-excited linear-prediction (CS-ACELP). G.729 offers a tradeoff: reduced overall bandwidth consumption with a slight reduction in voice quality. G.729 is ideally suited for the situation in which bandwidth is limited, such as in a WAN environment.

As you learned in previous chapters, several other features play a role in determining the bandwidth requirement of a voice call, including header compression, Layer 2 headers, and voice samples per packet. Voice Activation Detection (VAD) can also play a role in the bandwidth required by each call. VAD can be used to reduce the packet payload size by transmitting 2 bytes of payload during silent times rather than the full payload size. For example, the payload on a single G.711 packet using Cisco defaults is 160 bytes. VAD can reduce the size of the payload to 2 bytes during silent times in the conversations. Although VAD can offer bandwidth savings, Cisco recommends that VAD be disabled due to the possible voice-quality issues that it may induce. For the purposes of bandwidth engineering, VAD should not be taken into account.

Table 8-2 illustrates a few of the possible G.711 and G.729 bandwidth requirements.

Table 8-2 Bandwidth Requirements

Packet

IP/UDP/RTP Header Size

L2 Header Type

Second

Bandwidth per Call

G.711

160 bytes

40 bytes

Ethernet

14 bytes

50 pps

85.6 kbps

G.711

240 bytes

40 bytes

Ethernet

14 bytes

33 pps

77.6 kbps

G.711

160 bytes

40 bytes

MLPPP/FR

6 bytes

50 pps

82.4 kbps

G.711

160 bytes

2 bytes (cRTP)

MLPPP/FR

6 bytes

50 pps

67.2 kbps

G.729

20 bytes

40 bytes

Ethernet

14 bytes

50 pps

29.6 kbps

G.729

20 bytes

40 bytes

MLPPP/FR

6 bytes

50 pps

26.4 kbps

G.729

30 bytes

40 bytes

MLPPP/FR

6 bytes

33 pps

20 kbps

G.729

20 bytes

2 bytes (cRTP)

MLPPP/FR

6 bytes

50 pps

11.2 kbps

For DQOS test takers: These numbers are extracted from the DQOS course, so you can study those numbers. Note, however, that the numbers in the table and following examples do not include the L2 trailer overhead. Go to www.cisco.com, and search for "QoS SRND" for a document that provides some great background on QoS, and the bandwidth numbers that include data-link overhead.

For DQOS test takers: These numbers are extracted from the DQOS course, so you can study those numbers. Note, however, that the numbers in the table and following examples do not include the L2 trailer overhead. Go to www.cisco.com, and search for "QoS SRND" for a document that provides some great background on QoS, and the bandwidth numbers that include data-link overhead.

The formula used to calculate the bandwidth for this combination of factors is as follows:

Bandwidth per call = (Payload + IP/UDP/RTP + L2) * 8 * pps

For example, using G.729 @ 50 pps over Frame Relay without header compression results in the following calculation:

Bandwidth per call = (20 + 40 + 6) * 8 * 50 = 26.4 kbps

For example, using G.711 @ 50 pps over Ethernet without header compression results in the following calculation:

Bandwidth per call = (160 + 40 + 14) * 8 * 50 = 85.6 kbps

The elements in the bandwidth per call formula correspond to the following values:

• Payload—Payload size per packet depends on the codec selected and the number of voice samples in each packet. One voice sample represents 10 ms of speech. By default, Cisco includes two of these samples in each packet, transmitting 20 ms of speech in each packet. This means that there must be 50 packets per second to maintain a full second of voice conversation, as shown in the following:

20 ms * 50 pps = 1 second of voice conversation

After the number of samples per packet and packets per second has been determined, the payload size per packet is easily calculated by using the following formula:

Codec @ pps = (Codec payload bandwidth) / (Number of bits in a byte) / (Packets per second)

For example, the following shows a G.711 voice conversation using 50 pps:

G.711 @ 50 pps = 64 kbps / 8 bits / 50 pps = 160 bytes

For example, the following shows a G.711 voice conversation using 33 pps:

G.711 @ 33 pps = 64 kbps / 8 bits / 33 pps = 240 bytes

For example, the following shows a G.729 voice conversation using 50 pps:

G.729 @ 50 pps = 8 kbps / 8 bits / 50 pps = 20 bytes

For example, the following shows a G.729 voice conversation using 33 pps:

G.729 @ 33.334 pps = 8 kbps / 8 bits / 33.334 pps = 30 bytes

• IP/UDP/RTP headers—This is the combination of the IP header, UDP header, and RTP header overhead expressed in bytes. Without compression, this combination equals 40 bytes.

• Layer 2 header type—The Layer 2 transport technologies have the following header overheads:

— MLP over Frame Relay: 14 bytes

— MLP over ATM (AAL5): 5 bytes for every ATM cell + 20 bytes for the MLP and AAL5 encapsulation of the IP packet

Figure 8-4 illustrates the packet structure of the Layer 2 and IP/UDP/RTP headers and the payload for a voice packet.

Figure 8-4 Voice Packet Structure

Layer 2

IP

UDP

RTP

Payload of Speech Samples

Variable

Variable Size Based on Codec

Size

20

8

12

Selection and Number of

Based on

Bytes

Bytes

Bytes

Speech Samples Included

Layer 2

Protocol

• pps—The number of packets per second required to deliver a full second of a voice conversation. This value depends on the number of 10-ms samples within each packet. By default Cisco includes two 10-ms samples in each packet, transmitting 20 ms of sampled speech in each packet. If the number of samples per packet changes, the packets per second required to deliver a full second of voice conversation changes as well. If the packets per second increase, the overhead associated with the voice conversation increases, which requires additional bandwidth to deliver the same payload. Likewise, if the packets per second decrease, the overhead associated with the voice conversation decreases, which requires less bandwidth to deliver the same payload. The following calculations demonstrate the relationship between the packets per second and the samples included in each packet:

— 10 ms * 100 pps = 1 second of voice conversation

— 20 ms * 50 pps = 1 second of voice conversation

— 30 ms * 33 pps = 1 second of voice conversation

Armed with this information you can begin to build out bandwidth requirements based on the network infrastructure, codec, packet payload, and the number of simultaneous calls that need to be supported.

Figure 8-5 illustrates a small IP telephony network configured to use the G.711 codec @ 50 pps for all calls placed over the LAN; the G.729 codec @ 50 pps is used for all calls placed over the WAN.

In this example, RTP header compression and VAD are not in use and the Cisco default of 50 packets per second is assumed. A call from Host B phone to Host C phone across the switched LAN infrastructure consumes 85.6 kbps of bandwidth, as shown in the following equation:

A call placed from Host A phone across the WAN infrastructure to Remote A phone in this scenario requires 26.4 kbps, as shown in the following equation:

Assuming that you must allow 6 simultaneous calls across this WAN link at any given time, 158.4 kbps of WAN bandwidth is required to support the voice conversations, as shown in the following equation:

Figure 8-5 Bandwidth Considerations

Host Site

1 —1—-J 1

V □

■rl

CallManager Publisher

CallManager Publisher

CallManager Subscriber

Host Phone B Uses G.711 Across Ethernet to Call Host Phone C

Remote Site

Remote Phone A

Figure 8-5 Bandwidth Considerations

Host Site

Remote Site

Remote Phone A

CallManager Subscriber

Remote Phone C

Remote Phone C

Host Phone A Uses G.729 Across the Frame Relay Cloud to Call Remote Phone A.

Assuming that you must provide for a guaranteed minimum of 256 kbps for data traffic, the total circuit bandwidth requirements can be derived from the following formula:

(Number of calls desired * bandwidth per call) + (Total data requirements)

Examining circuit speeds available today, a 512-kbps link can be used for this IP telephony network to meet the assumed voice and data requirements for 414.4 kbps. The remaining 97.6 kbps can be used for additional overhead, such as routing protocols.

Table 8-3 illustrates the relationship between codec, header compression, number of simultaneous calls, and the minimum bandwidth required for data traffic. Although the number of simultaneous calls, packet payload, and data requirements remained constant in this example, the codec selection and header compression varied the total circuit bandwidth requirements significantly.

Table 8-3 Impacting the Total Bandwidth Requirements

Codec

Compression

Bandwidth per Call

Maximum Number of Calls Required

Voice

Bandwidth

Minimum

Bandwidth

Circuit

Bandwidth

G.729

No

26.4 kbps

6

158.4 kbps

256 kbps

414.4 kbps

512 kbps

G.729

RTP header compression

11.2 kbps

6

66.6 kbps

256 kbps

322.6 kbps

512 kbps

G.711

No

82.4 kbps

6

494.4 kbps

256 kbps

750.4 kbps

768 kbps

G.711

RTP header compression

67.2 kbps

6

403.2 kbps

256 kbps

659.2 kbps

768 kbps

When you have a clear understanding of the bandwidth required for supporting the addition of voice on your packet network, you can begin to design the proper CAC mechanisms for your converged network.

Advance SEO Techniques

Advance SEO Techniques

Turbocharge Your Traffic And Profits On Auto-Pilot. Would you like to watch visitors flood into your websites by the 1,000s, without expensive advertising or promotions? The fact is, there ARE people with websites doing exactly that right now. How is that possible, you ask? The answer is Advanced SEO Techniques.

Get My Free Ebook


Post a comment