Technical data

DATA CENTER and CAMPUS NETWORKS DEPLOYMENT GUIDE
Deploying Brocade Networks with Microsoft Lync Server 2010 9 of 52
• Processing delay includes the time required to collect a frame of voice samples before processing by
the speech encoder can occur—the actual process of encoding, encrypting if appropriate, packetizing
for transmission—and the corresponding reverse process on the receiving end, including the jitter
buffer used to compensate for varying packet arriving delay on the receiving end. The complete end-to-
end processing delay is often in the 60 ms to 120 ms range, when all of the contributing factors are
taken into account. The processing delay is essentially within a fixed range determined by the vendor’s
technology and implementation choices. Encoding and decoding might be repeated several times;
however, if there is any inline transcoding from one codec to another—for example, for hand-off
between networks—then accumulated processing delay can become disruptive.
• Serialization delay is a fixed delay required to clock a voice or data frame onto a network interface,
placing the bits onto the wire for transmission. The delay varies based on the clocking speed of the
interface. A lower-speed circuit (such as a modem interface or smaller transmission circuit) has a
higher serialization delay than a higher-speed circuit. The delay can be quite significant on low-speed
links and occurs on every single link of a multihop network.
• Network delay is mostly caused by inspecting, queuing, and buffering of packets, which can occur at
traffic shaping buffers (such as “leaky bucket” buffers), which are sometimes encountered at various
network ingress points or at various router hops encountered by the packet along the way. Network
delay on the Internet generally averages less than 40 ms when there is no major congestion. Typically,
an easy way to spot congestion is when network delays start to increase. It is good practice to create
alarms and alerts to detect such issues, so that you can quickly resolve the problem. Modernization of
routers has contributed to reducing this delay over time.
• Propagation delay is the distance traveled by the packet, divided by the speed of signal propagation
(that is, the speed of light). Propagation delay on transcontinental routes is relatively small--typically
less than 40 ms—but propagation delay across complex intercontinental paths can be much greater.
This is especially true when satellite circuits are involved or on very long routes, such as Australia to
South Africa via Europe, which might incur up to 500 ms of one-way propagation delay.
The sum of these four delay components creates the total delay. The ITU-T has recommended 150 ms total one-way
delay (including endpoints) as the upper limit for “excellent” voice quality. Longer delays can be disruptive to the
conversation, with the risk of talkover effects and echo. When the one-way delay exceeds 250 ms, it is likely that
talkers will step over each other’s speech, which is known as step-over.
In the event of a transcontinental route with well-sized links, the total delay in non-congested conditions might be
70 ms (processing), plus 10 ms (serialization), plus 30 ms (network), plus 40 ms (propagation), which equals 150
ms total. Therefore, IP telephony calls frequently function where even small incremental delays could impact the
voice quality.
Network delay is the one component over which the system administrator has the most control. Network delay can
be reduced through a variety of network engineering means. However, the first priority of network delay engineering
is often avoidance of spikes and limitation of variability (that is, jitter) due to congestion—ahead of reduction in
normal delay. Of all the delay components, queuing at router hops is the most variable and unpredictable
component of overall delay, especially in situations of congestion. This makes it one of the areas in which Quality of
Service (QoS) techniques are most frequently used.
As network demands increase from voice and video, it is important to have network switches and routers that are
capable of scaling for these demands. Brocade switches give you the ability to scale to newer 10 Gigabit Ethernet
(GbE) technology, setting traffic rate limiting, and enabling end-to-end QoS. Network administrators need to carefully
assess their network trends over a period of time so that they can be sure of having adequate bandwidth during
peak times. In addition, Brocade IronView
®
Network Manager (INM) software can assist a network administrator in
analyzing the network and identifying hot spots.