Traffic engineering

The TDM usage rates of 130.0 Erlangs for Site 1 and 46.7 Erlangs for Site 2 can both be easily handled by the TDM facilities of a single PN, which is capable of carrying up to 200 Erlangs of TDM traffic at a P001 GOS. Therefore, the assumption that all calls between two circuit-switched endpoints is intra-Port Network is valid. If one PN was insufficient to support the TDM usage in one of the sites, the calculations would have been repeated under the assumption of two PNs. If a pair of PNs was still insufficient, the number would continually be incremented until there were enough port networks to handle the TDM usage in that particular site. Finally, the TDM resources on the two G350 Media Gateways are easily sufficient for supporting the 17.9 Erlangs of TDM traffic in Site 3.

Note:

The more PNs, the more inter-Port Network calls there are, and hence more TDM usage, since each interport network call requires resources in each PN that is involved in each call.

Next, the media processing resources must be considered. Since there are some fundamental differences between the TN2302AP and the TN2602AP media processors, they will be discussed separately, beginning with the TN2302AP.

Each TN2302AP IP Media Processor circuit pack (or Media Gateway VoIP Media Module) can support only a finite number of simultaneous calls. However, the exact number that can be supported varies according to the codecs of the calls to be supported. In general, compressed calls (for example, G.729 codec) require twice as many media processing resources as uncompressed calls (for example, G.711 codec). Also, calls utilizing AES media encryption require approximately 25% more media processing resources than unencrypted calls.

A TN2302AP circuit pack (or a MG VoIP Media Module) can support both compressed and uncompressed calls, as well as both encrypted and unencrypted calls, all simultaneously. Therefore, the general model for sizing the media processing resources is very complex. The model is a “batch arrival and service” model, and the details are beyond the scope of this document.

In practice, a fairly common strategy is to use an uncompressed codec for intrasite calls, and a compressed codec for intersite calls. This is due to the trade-off between bandwidth savings, increased media processing costs, and voice quality for compressed calls. If a private LAN is used for intrasite calls, bandwidth usage is of less concern than media processing cost and voice quality. However, for intersite calls, especially over a public WAN, the bandwidth savings offered by the use of compression outweighs the extra processing costs and slight degradation of voice quality.

Recall that any usage that is expressed in Erlangs represents the average number of busy servers at any given time. For the total media processing usages provided at the bottom of Table 34, a “server” can be thought of as the set of media processing resources that is necessary to support a single bidirectional media stream through a media processing circuit pack. Consider the total of 89.4 Erlangs of media processing usage in Site 1. This usage consists of 76.8 Erlangs of intrasite usage, and 12.6 Erlangs of intersite usage. Assume that an uncompressed codec is used for the intrasite calls, and a compressed codec is used for the intersite calls. Since each compressed call requires twice as many media processing resources as each uncompressed call, the 12.6 Erlangs must be counted twice. Therefore, the media processing load is actually 76.8 + (2 x 12.6) = 102.0 Erlangs. Similarly, the total media

208 Avaya Application Solutions IP Telephony Deployment Guide

Page 208
Image 208
Avaya 555-245-600 manual Traffic engineering