Delay means the period of time a packet takes to traverse from one point in the network to another. Furthermore the time required for a packet to be passed from the source to its destination is called end-to-end delay. The components that contribute to end-to-end delay may be fixed or variable, and the following is a discussion of the various sources of delay for an IP packet.
A) Sender end 1. Coding Delay
Codecs are used to convert analog voice (or video) signals to digital data. They also perform voice compression to reduce the bandwidth required over IP networks. This conversion from analog to digital and then compression introduces delays in the codec at the sender. Different coding can result in different codec delay, and higher compression is achieved at the price of longer delays. Table 2-2 gives encoding and decoding delays for several typical voice codecs standardized by the International Telecommunication Union (ITU). As for video, the codec delay may be in the order of tens of milliseconds. Coding standard Compression algorithm Bit rate (Kbps) Encoding delay (ms) Typical decoding delay (ms) Total coding delay (ms) G.711 PCM 64 0 0 0 G.729 CS-ACELP 8 15 7.5 22.5 G.723.1 ACELP 5.3/6.3 37.5 18.75 56.35
Table 2-2 Encoding and decoding delays for voice packet [Has00]
2. Packetization Delay
The encoded bit-stream generated is then packetized, and this incurs a packetization delay. This packetization delay equals the IP payload size divided by the source
Chapter 2 Introduction to Real-time IP Traffic and Active Queue Management
information rate. 3. Serialization Delay
Serialization delay is the time required to transmit all of the packet’s bits onto the link. It depends on the packet’s length and the transmission rate of the link. With higher line speeds, serialization delay can be greatly reduced. The serialization delay is often of the order of microseconds to milliseconds in practice.
B) Network
4. Propagation Delay
The time required to propagate from the beginning of the link to the next node is the propagation delay. The propagation delay is related to the physical medium of the link and its propagation speed, e.g. fiber optics, twisted –pair copper wire, and the distance between two nodes. The propagation delay is the distance between two nodes divided by the propagation speed of the link. In wide-area networks, propagation delay is typically of the order of milliseconds.
5. Queuing Delay and Processing Delay
Queuing delay is the time the packet stays in the node, and it comprises the processing delay and the queuing delay. For the processing delay, time is required for the router to determine the next hop location according to the packet header and the routing mechanisms. This processing time can be much reduced by using fast forwarding schemes like ATM (Asynchronous Transfer Mode) or MPLS (Multiprotocol Label Switching).
Queuing delay occurs when the queue service rate is not fast enough to serve all the incoming packets from different sources using this queue, therefore, the packets accumulate in the queue and wait for service. The queuing behaviour depends on the traffic pattern, i.e. the statistical nature of the arrival process. The queuing delay at a node can vary significantly from packet to packet due to the randomness of the incoming aggregate packet arrival process. So the characteristics of queuing delay typically are given by statistical measures, such as average queuing delay, variance of
Chapter 2 Introduction to Real-time IP Traffic and Active Queue Management
queuing delay exceeds some specific value. C) Receiver end
6. Playback Delay
Due to variable queuing delays, it is impossible for all the packets in a stream to arrive at the receiver with the same inter-arrival time. To allow for variable packet arrival times and to achieve a steady stream of packets, the receiver holds packets in a buffer for a while before playing them. This holding time causes further delay for the application.
7. Decoding and De-packetization Delay
This is the time required to re-assemble the packets to the original form of the signal. They are the counterparts to coding and packetization at the sender, as shown in Table 2-2.
8. Summary of delays
Assuming that a packet starts from the source, passes through a series of routers, and ends in the destination, the packet will suffer several types of delays at each node along the path. These delays include coding delay, packetization delay, transmission delay, queuing delay, propagation delay, playback delay, decoding delay and de-packetization delay. The end-to-end delay is the sum of the delays experienced at each hop on the way to its destination.
Delay introduced above can be divided into two components, a variable component and a fixed component. The only variable part of delay is the time spent in the queues of the network nodes on the transmission path. They are very hard to predict and depend heavily on the current network load. So when network load is stable, investigating how to reduce this variable part of delay for real-time delay sensitive traffic is important for QoS research; the fixed part of delay is independent of network conditions and depends primarily on the number and the performance of network nodes on the transmission path and the link capacity, etc. The fixed component of delay includes all the other delays except queuing delays. These fixed
Chapter 2 Introduction to Real-time IP Traffic and Active Queue Management