• No se han encontrado resultados

Las prácticas del modelo de politización solidaria alternativa

b eRnabé m alacalza **

2. Los modelos de cooperación en Haití después del terremoto

2.4. Las prácticas del modelo de politización solidaria alternativa

Contents

This section contains information on the following topics:

"Introduction" (page 84)

"Performance criteria" (page 85)

"Network performance evaluation overview" (page 86)

"Network performance measurement tools" (page 91)

"Network availability" (page 91)

"Bandwidth" (page 93)

"Available Bandwidth" (page 93)

"Guaranteed Bandwidth" (page 94)

"Queueing" (page 94)

"Calculating per-call bandwidth use" (page 94)

"Silence Suppression engineering considerations" (page 99)

"Estimate network loading caused by VoIP traffic" (page 100)

"Route Link Traffic estimation" (page 104)

"Enough capacity" (page 105)

"Insufficient link capacity" (page 106)

"Other intranet resource considerations" (page 107)

"Delay" (page 107)

"Effects of delay on voice quality" (page 109)

"Components of delay" (page 109)

"Measuring end-to-end network delay" (page 112)

"Adjusting PING statistics" (page 114)

"Other measurement considerations" (page 115)

"Reducing delays" (page 115)

"Recording routes" (page 116)

"Routing issues" (page 116)

"Jitter" (page 116)

"Jitter buffers" (page 118)

"Late packets" (page 118)

"Adjusting jitter buffer size" (page 118)

"Jitter measurement tools" (page 120)

"Packet loss" (page 120)

"Physical medium loss" (page 120)

"Congestion loss" (page 120)

"Measuring end-to-end packet loss" (page 122)

"Packet Loss Concealment" (page 122)

"Reducing packet loss" (page 123)

"Network delay and packet loss evaluation example" (page 123)

"Estimate voice quality" (page 124)

"Sample scenarios" (page 128)

"Does the intranet provide expected voice quality?" (page 129)

Introduction

To create a VoIP-grade network, certain QoS standards for basic network elements must be met. The following QoS parameters can be measured and monitored to determine if desired service levels have been obtained:

network availability

bandwidth

delay

jitter

packet loss

These QoS parameters and mechanisms affect the application’s or end-user’s Quality of Experience (QoE). These QoS parameters apply to any IP (Internet Protocol) network carrying VoIP traffic, including LANs, campus-wide networks, and WANs.

Performance criteria

This section describes criteria for achieving excellent voice quality. The network should meet these specifications.

End-to-end packet delay: Packet delay is the point-to-point, one-way delay between the time a packet is sent to the time it is received at the remote end. It is comprised of delays at the Voice Gateway Media Card, Internet Telephone, and the IP network. To minimize delays, the IP Telephony node and Internet Telephone must be located to minimize the number of hops to the network backbone or WAN.

Note: Nortel recommends an end-to-end delay of <= 50 ms on the IP network to ensure good voice quality. This does not include the built-in delay of the Voice Gateway Media Card and IP Phone.

End-to-end packet loss: Packet loss is the percentage of packets sent that do not arrive at their destination. Transmission equipment problems, packet delay, and network congestion cause packet loss. In voice conversation, packet loss appears as gaps in the conversation.

Sporadic loss of a few packets can be more tolerable than infrequent loss of a large number of packets clustered together.

Note: For high-quality voice transmission, the long-term average packet loss between the IP Phones and the Voice Gateway Media Card TLAN network interface must be < 1%, and the short-term packet loss must not exceed 5% in any 10-second interval.

Recommendation

To achieve excellent voice quality, Nortel strongly recommends using G.711 CODEC with the following configuration:

end-to end delay less than 150 ms one way (network delay + packetization delay + jitter buffer delay < 150). See"IP expansion link Packet Delay Variation jitter buffer" (page 163).

packet loss less than 0.5% (approaching 0%)

maximum jitter buffer setting for IP Phone as low as possible (maximum 100 ms)

Packet loss on the ELAN network interface can cause:

— communication problems between the Call Server and the Voice Gateway Media Cards

— lost SNMP alarms

— incorrect status information on the TM console

Note: Since the ELAN network is a Layer 2 Switched LAN, the packet loss must be zero. If packet loss is experienced, its source must be investigated and eliminated. For reliable signaling communication on the ELAN network interface, the packet loss must be < 1%.

Network performance evaluation overview

There are two main objectives when dealing with the QoS issue in an IP network:

1. Predict the expected QoS.

2. Evaluate the QoS after integrating VoIP traffic into the intranet.

The process for either case is similar—the first is without VoIP traffic, and the second is with VoIP traffic. The differences are discussed in this section.

This process assumes that the PING program is available on a PC, or some network management tool is available to collect delay and loss data and to access the LAN that connects to the router to the intranet.

Procedure 5

Evaluating network performance: overview Step Action

1 Use PING or an equivalent tool to collect round-trip delay (in ms) and loss (in%) data.

2 Divide the delay (determined instep 1) by 2 to approximate one-way delay. Add 93 ms to adjust for ITG processing and buffering time.

3 Use a QoS chart, orTable 21 "QoS levels" (page 127), to predict the QoS categories: Excellent, Good, Fair or Poor.

4 If a customer wants to manage the QoS in a more detailed fashion, re-balance the values of delay compared to loss by adjusting system parameters, such as preferred CODEC, payload size, and routing algorithm, to move resulting QoS among different categories.

5 If the QoS objective is met, repeat the process periodically to make sure the required QoS is maintained.

—End—

Set QoS expectations

The users of corporate voice and data services expect these services to meet some perceived Quality of Service (QoS) which in turn influences network design. The goal is to design and allocate enough resources in the network to meet users’ needs. QoS metrics or parameters are what quantifies the needs of the "user" of the "service".

In the context of a Meridian 1, CS 1000S, and CS 1000M system,Figure 19 "QoS parameters" (page 87)shows the relationship between users and services.

Figure 19 QoS parameters

InFigure 19 "QoS parameters" (page 87), there are two interfaces to consider.

The Meridian 1, including the IP Trunk 3.0 (or later) nodes, interfaces with the end users. Voice services offered by the Meridian 1 must meet user-oriented QoS objectives.

The IP Trunk 3.0 (or later) nodes interface with the intranet. The service provided by the intranet is "best-effort delivery of IP packets", not

"guarantee QoS for real-time voice transport." IP Trunk 3.0 (or later)

The QoS level is a user-oriented QoS metric which takes on one of four settings – Excellent, Good, Fair, or Poor – indicating the quality of voice service. IP Trunk 3.0 (or later) periodically calculates the prevailing QoS level per site pair, based on its measurement of the following:

one-way delay

packet loss

CODEC

Recommendation

Nortel strongly recommends that G.711 CODEC be used over high-bandwidth connections, and used any time that call quality is the highest priority. Where call quality is the highest priority, sufficient bandwidth must be provided for the VoIP application. The Best Quality (BQ) CODEC is usually chosen and configured as G.711 within the zone configuration (intrazone).

Use G.729 CODEC to compress voice traffic over low-bandwidth connections when bandwidth considerations take precedence over call quality. The Best Bandwidth (BB) CODEC is usually chosen and set to G.729A or G.729AB between zones (interzone).

CODEC details are then configured on the Signaling Server through TM or CS 1000 Element Manager.

Figure 20 "QoS levels with G.729A/AB CODEC" (page 89),Figure 21 "QoS level with G.711 CODEC" (page 90), andFigure 22 "QoS level with G.723 CODEC" (page 90)are derived from the ITU-T G.107 Transmission Rating Model. These diagrams show the operating regions in terms of one-way delay and packet loss for each CODEC. Note that among the CODECs, G.711 A-law/G.711 mu-law delivers the best quality for a given intranet QoS, followed by G.729AB, G.723.1 6.4 kbps, and G.723.1 5.3 kbps. These graphs determine the delay and error budget for the underlying intranet so it delivers a required quality of voice service.

Fax is more susceptible to packet loss than is the human ear, in that quality starts to degrade when packet loss exceeds 4%. Nortel recommends that fax services be supported with IP Trunk 3.0 (or later) operating at the Excellent or Good QoS level. Avoid offering fax services between two sites that can guarantee no better than a Fair or Poor QoS level.

G.729AB CODEC

The G.729 uses less bandwidth than the G.711. If minimizing bandwidth demand is a priority, and the customer is willing to accept lesser voice quality, a G.729AB CODEC can be used.

Extreme care must be taken in the network design if using the G.729AB CODEC. The G.729AB CODEC has the same requirements as the G.711 CODEC.

Figure 20 "QoS levels with G.729A/AB CODEC" (page 89)shows the QoS levels with a G.729A/AB CODEC.

Figure 20

QoS levels with G.729A/AB CODEC

G.711 CODEC

G.711 is the recommended CODEC.

Figure 21 "QoS level with G.711 CODEC" (page 90)shows the QoS levels with a G.711 CODEC.

Figure 21

QoS level with G.711 CODEC

G.723 CODEC

Figure 22 "QoS level with G.723 CODEC" (page 90)shows the QoS levels with a G.723 CODEC.

Figure 22

QoS level with G.723 CODEC