Resource sizing
Example 7: LAN bandwidth
In Example 6: IP bandwidth considerations, the total IP LAN bandwidth usage for each site was calculated, and expressed in Erlangs at the bottom of Table 39. Specifically, the total LAN bandwidth usage in Site 1 is 123.5 Erlangs, in Site 2 is 56.6 Erlangs, and in Site 3 is 29.2 Erlangs. This implies that the average number of bidirectional media streams that are simultaneously in use at any given time in Site 1 is 123.5. Analogous statements can also be made regarding Sites 2 and 3.
Every media stream across the IP LAN in any of the three sites is assumed to use the uncompressed G.711 codec, since bandwidth is relatively inexpensive within a private LAN, as opposed to a public WAN. Assume, for the sake of this example, a standard IP packet size of 20 ms. So for the G.711 codec, Table 42 indicates that each media stream consumes 87.2 kbps of IP LAN bandwidth. It may be tempting at this point to simply multiply 87.2 kbps by 123.5 simultaneous bidirectional media streams, to arrive at the estimate for the overall LAN bandwidth needed for Site 1. However, 123.5 is merely the average number of simultaneous media streams, and approximately half of the time, there are at least 124 simultaneous media streams in use.
In this example, suppose that the goal is to supply enough bandwidth to adequately support the media streams at least 99.9% of the time. The standard
Table 43: IP LAN bandwidth requirements in each direction, for Example 7: LAN bandwidth
Resource | Site 1 | Site 2 | Site 3 |
| (Atlanta) | (Boston) | (Cleveland) |
|
|
|
|
Simultaneous media streams for | 158 | 81 | 47 |
“P001” |
|
|
|
|
|
|
|
LAN bandwidth (Mbps) | 13.8 | 7.1 | 4.1 |
|
|
|
|
In Table 43, the number of simultaneous media streams for “P001” represents the 99.9th percentile for the number of simultaneous unidirectional streams, as determined by applying the standard
A slight variation of the procedure that was used to determine LAN bandwidth in Example 7: LAN bandwidth can be used to determine WAN bandwidth. Using compressed RTP (cRTP) is a means to conserve bandwidth. Specifically, the use of cRTP reduces the overhead due to IP, UDP, and RTP from 40 bytes to between 2 and 4 bytes (4 bytes are assumed for this example). Using the PPP overhead of 7 bytes (which would vary if ATM, HDLC, or Frame Relay were used) implies a total overhead of 11 bytes (88 bits) in this example. This implies the following
Issue 6 January 2008 221
