Page 166 - DCAP406_DCAP_207_Computer Networks

Unit 10: Network Layer in the Internet

10.3.2 QoS Concepts Notes

Congestion Management: The bursty nature of data traffic, sometimes bounds to increase the
amount of traffic more than the speed of a link. In such a situation, QoS enables a router to put
packets into different queues and service certain queues more often based on priority rather
than buffer traffic in a single queue and let the first packet in be the first packet out. Such issues
are incorporated in congestion-management tools to handle. Thus, the congestion management
tool may include priority queuing, custom queuing, weighted fair queuing, etc.
Queue Management: The queues in a buffer may fill and overflow. A packet will be dropped if
a queue is full and router cannot prevent this packet from being dropped even if it is a high-
priority packet. This is referred as tail drop. It could be prevented either by ensuring that the
queue does not fill and provide room for high-priority packets or allow some rule for dropping
packets with lower priority before dropping higher-priority packets. A mechanism called
weighted early random detect perform both of these functions.
Link Efficiency: Sometimes, the low-speed links are bottlenecks for smaller packets. The
serialization delay caused by the large packets force the smaller packets to wait longer. The
serialization delay is the time taken to put a packet on the link. The serialization delay (For
example, the serialization delay for a 2400-byte packet on a 56-kbps link will be 343 milliseconds)
will make a voice packet, which is behind it in queue to delay enormously before the packet left
the router, a situation, which is not desirable for voice packets. The link fragmentation and
interleave process segment large packet into smaller packets interleaving the voice packet.

Elimination of overhead bits: The efficiency could also be improved by eliminating too many
overhead bits. For example, RTP headers have a 40-byte header and a payload of as little as 20
bytes. In such a case, the overhead is twice that of the payload. Some compression technique may
be applied to reduce the header to a more manageable size.
Traffic shaping and policing: Shaping is used to prevent the overflow problem in buffers by
limiting the full bandwidth potential of the packets of applications. Sometimes, in many network
topologies that has a high-bandwidth link connected with a low-bandwidth link in remote sites
may overflow low bandwidth link. Hence, shaping is used to provide the traffic flow from high
bandwidth link closer to the low bandwidth link to avoid the overflow of the low bandwidth
link. Policing is used to discard the traffic that exceeds the configured rate but in case of shaping
it is buffered.

End-to-end QoS Levels

It refers to the capability of a network to deliver service needed by specific network traffic from
end to end or edge to edge under network constraints like bandwidth, delay, jitter, loss
characteristics, etc. These factors describe how tightly the end-to-end service performs. The QoS
involves a policy framework or set of rules that designate an action. The policy framework
provides a particular service to particular client, application and schedule. Three basic levels of
end-to-end QoS can be provided across a heterogeneous network. They are integrated service,
differentiated service and inbound admission service types.
Performance Limits: The performance limits also considers the token bucket limits and bandwidth
limits together to guarantee packet delivery in outbound bandwidth policies for the integrated
and differentiated service.
Token Bucket Size: When an application is sending information faster than the server sends the
data out of the network, the buffer will fill up. To avoid such situations, the token bucket size is
applied to determine the amount of information a server can process at any given time. A packet

LOVELY PROFESSIONAL UNIVERSITY 159