Table of Contents
Advertisement
Chapter 5: Operating Modes and Configuration
Packet Flow
Packet Order
802.1p Class of Service (CoS), and DiffServ when appropriate, levels are configurable to ensure timely
delivery of TDM packets through the LAN.
While traveling across packet switched networks, received packets may arrive out of sequence for a variety
of reasons. E3Switch gateways will use sequence numbers, present in the packets, to restore original packet
order.
Receive Jitter Buffer Depth and Latency
While traveling across packet switched networks, received packets may exhibit slight variations in traversal
time. The function of the configured receive-buffer-depth is to absorb this jitter and ensure that there is
always data available to present to the outgoing, constant-bit-rate, TDM circuit when an incoming LAN
packet is delayed. Larger configured buffer depths result in greater latency, so it is desirable to determine
the minimum required value for this setting. The amount of LAN packet jitter in the system can be
determined at the HTTP web management status page by examining the "Min⁄Cur⁄Max Pkts Waiting" and
"Jitter Buffer Underflow" counters. The "Jitter Buffer Depth" value should be adjusted at the HTTP web
management settings page until no "Jitter Buffer Underflow" errors accumulate and the "Min Pkts Waiting"
level is at least one. The typical latency incurred per, 1000-byte, LAN packet of jitter buffer depth in a full-
rate TDM circuit is 0.2ms. Configuring a, rather deep, 5-packet buffer would introduce 1ms of delay into
the system. For fractional TDM circuits with large TDM Frames per Packet settings described below, the
latency introduced is the product of the "Jitter Buffer Depth" and the latency per LAN packet, which can
become too large if care is not taken.
TDM Frames per Packet and Latency
For the following discussion, if the incoming TDM is unframed, the term "frame" simply refers to the
number of TDM bits that would constitute a frame if framing were in effect.
Each LAN packet requires approximately 50-bytes of overhead for transmission. Accordingly, on limited
bandwidth LANs, it is desirable to consolidate as many complete, TDM frames as possible into each LAN
packet. At full T3 TDM speed, this results in two frames per packet and a latency of approximately 0.2ms.
At full E3 speed, this results in seven frames per packet and a latency of approximately 0.3ms.
When fractional TDM is in effect, as little as 4-bytes per frame may be in use. In this case, transmitting
only two frames per LAN packet would still result in a 0.2ms latency, while waiting for each TDM frame to
arrive; however, only 8 bytes of each 64-byte minimum LAN packet, and thus 13% of the required LAN
bandwidth, would be in use for actual TDM data. This efficiency level may be too low in limited-
bandwidth applications such as satellite links. Increasing the number of frames per packets in this case will
increase the efficiency of the transmission but at a cost of approximately 0.1ms latency per T3 frame or
0.05ms latency per E3 frame.
Latency in milliseconds is approximately equal to 1000 x [frames per LAN packet] x [bits per full TDM
frame] / [full TDM bitrate] – where [bits per full TDM frame] is 4760 for T3 and 1536 for E3; full TDM
bitrate is 44,736,000 for T3 and 33,368,000 for E3. To achieve 97% efficiency of a 10Mbit/s satellite link
in a fractional E3 environment, 9.7Mbit/s of E3 data / 33.368Mbit full E3 rate x 1536 = 446 bits / 8 = 55
bytes per E3 frame in use, allowing about 26 frames per LAN packet. The latency incurred if this
maximum efficiency packing were in use would be 1000 x 26 x 1536 / 33,368,000 = 1.2ms. This single-
packet latency must now be multiplied by the typical or maximum jitter buffer depth described in the prior
section of this document to estimate the entire latency of the system.
12
Advertisement
Table of Contents
