The idea
The Internet was built to deliver data on a best-effort basis. That is fine for email and file transfers, which can wait. It is not fine for a phone call, where a late packet is a useless packet. Quality of service is the vocabulary for saying what a particular flow needs from the network, and the mechanisms for giving it that.
How it works
What network QoS is
The lecture gives two descriptions of network QoS:
- A description or measurement of the overall performance of the network.
- A set of techniques and mechanisms that guarantee the performance of the network so that it delivers predictable service.
It adds that the Internet was originally designed for best-effort service, which carries elastic traffic, and that QoS may be loosely defined as something a flow seeks to attain. The topics of the lecture are flow characteristics, QoS metrics, QoS principles, traffic management and resource management.
How it works
The four flow characteristics
Traditionally, a flow is described by four characteristics.
Reliability. Lack of reliability means losing a packet or an acknowledgment, which entails retransmission. Applications differ. Reliability is more important for email, file transfer and web browsing, and less important for telephony and audio/video conferencing.
Delay. Source-to-destination delay consists of the sum of delays at routers, the propagation delay in the media, and set-up mechanisms. Tolerance is application specific. Telephony, audio/video conferencing and remote log-in tolerate minimal delay. File transfer and e-mail tolerate a lot.
Jitter. The variation of delay for packets in the same flow. Low jitter means the differences are small, high jitter means they are large. Variation of delay is not acceptable for audio and video applications, except with sufficient application buffering.
Bandwidth (capacity). The maximum rate at which data can be transferred
over a given link, measured in bits per second (bps). Requirements are
application specific and flow specific.
How it works
Network delay
Network delay has four components:
- Transmission delay, packet length divided by transmission rate.
- Propagation delay, distance divided by propagation speed.
- Processing delay, the time required to process a packet in a router or a destination host.
- Queuing delay, the time a packet waits in input and output queues in a router.
Propagation delay is introduced earlier in Signals.
Worked example
Transmission delay for one packet
Given (our own illustration, the lecture gives no numbers here): packet length
8000 bits, transmission rate2,000,000 bps.Formula: transmission delay = packet length / transmission rate.
Substitute:
8000 bits / 2,000,000 bps.Result:
0.004 s=4 ms.
Answer0.004 s, which is 4 ms.
Where marks get lost
Jitter is not delay
A flow where every packet takes exactly 200 ms has a large delay and no jitter. A flow where packets take 20, 120 and 40 ms has a small average delay and high jitter. The lecture’s point is that audio and video suffer from the variation, and that buffering in the application can hide it.
In the exam
- Four characteristics: reliability, delay, jitter, bandwidth. Name an application that is sensitive to each and one that is not.
- Four delay components and the two formulas.
- Jitter needs the words “variation of delay” and “same flow”.
- Bandwidth is a maximum rate in
bps. Throughput is a different thing, covered in QoS metrics.
Check yourself
- QoS describes network performance and the mechanisms that make it predictable.
- A flow is characterised by reliability, delay, jitter and bandwidth.
- Delay is transmission, propagation, processing and queuing.
- Jitter is delay variation within a flow. Buffering hides it for audio and video.