Afar NX-2E1, Operator'S Manual

Page: 23 / 75

NetCrossing Gateway NX-2E1/T1 Operator’s Manual
2-5
Table 2.4 – WAN data rates (T1 mode)
Bridge Route
Payload Time slots Pkt rate Pkt size
(bytes)
Data rate
(Kbps)
Pkt size
(bytes)
Data rate
(Kbps)
Unframed N/A 250 804 1608 824 1648
24 333 608 1621 628 1675
16 333 416 1109 436 1163
8 333 224 597 244 651
Framed 4 333 128 341 148 395
2 333 80 213 100 267
1 333 56 149 76 203
At the receiving side, the gateway removes the packet encapsulation, and queues the serial data
portion of the packet to be sent out over the appropriate serial TDM port. In fractional mode the
empty time slots are filled with a configurable idle code.
2.3.3 Jitter Buffer and Link Latency
For each TDM line a gateway transmits over the WAN a packet containing serial data every 4.0
milliseconds (in T1 mode it is every 3.0 milliseconds). If the WAN provided an instantaneous
delivery of these packets, the receiving gateway could be sending out, over the serial port, the last bit
from the previous packet, when the next packet would arrive providing the next 4.0 ms worth of serial
data. In this case the latency in the end to end serial link would be exactly 4.0 ms due to the store and
forward delay.
However, the WAN network introduces its own delay, and worst, the delay introduced is not constant
for every packet. The variation in the packet delivery time across the WAN is called
jitter . When a
serial packet is delivered late, the receiving gateway might run out of serial data causing an underrun
error. In order to avoid these errors, the gateways store a certain amount of serial data such that the
serial port always has data, even when a packet is delivered late. The size of this
jitter buffer is an
important parameter of the gateway configuration. It must be made large enough to absorb the worst
case jitter, but the link latency also increases by the same amount.
Sometimes the jitter in the WAN network is hard to predict or varies over time. If the WAN traffic is
light when you establish the link, you might set the jitter buffer to a low value, only to find that later
in the day, as the WAN traffic peaks, a lot of underrun errors occur.