Traffic Control & QOS

What is Engineering ?
It is step by step procedure of

Design Something
Handle Something
Maintain Something
Apply Something
Invent Something

What is Traffic ?
Mess of something
Flow in the path etc.

What is Network ?
It is connection of related things say for ex.
mobile clusters, connection of PCs within some
range say global range i.e. internet

What is Network Traffic Engineering ?

It is designing, handling, maintaining the flow in
the path of a network....
Traffic Engineering (TE) is one efficient approach
to
handle traffic changes in a network

Network Topologies :
Mesh topology
Star topology
Ring topology

in simple words, it is physical arrangement of

PCs
and network elements in a network

Introduction :
Various applications deployed over the Internet
are such as peer-to-peer, video on demand etc.
The traffic generated by such applications causes
unpredictable large changes in traffic demand
The fixed routes can be designed so as to handle
any possible traffic changes, but this approach
requires more than double the bandwidth required
for the routes suitable to the traffic at each time

What is option?
Traffic Engineering (TE), efficient way of handling
such traffic changes
Deployed a server called Path Computation Element
(PCE)
Collects the information of traffic amounts between all
nodes periodically then recalculates the routes within the
network so as to accommodate all the current traffic
without any congestion according to the collected traffic
information

Microsoft Windows [Version 6.1.7600]

Copyright (c) 2009 Microsoft Corporation. All rights reserved.
C:\Users\Vaibhav>tracert www.google.com
Tracing route to www.google.com [173.194.38.146]
over a maximum of 30 hops:
1 <1 ms <1 ms
2 36 ms 37 ms
3 37 ms 38 ms
4 134 ms 136 ms
[115.113.128.17]
5 141 ms 135 ms
[115.113.165.98]
6 141 ms 138 ms
7 165 ms 165 ms
8 282 ms 281 ms
9 288 ms 282 ms
10 282 ms 282 ms
Trace complete.

<1 ms 192.168.1.1
37 ms 117.223.32.1
37 ms 218.248.173.214
139 ms 115.113.128.17.static-mumbai.vsnl.net.in
135 ms 115.113.165.98.static-mumbai.vsnl.net.in
140 ms
168 ms
282 ms
282 ms
282 ms

72.14.232.202
66.249.94.39
66.249.94.72
72.14.233.79
sin04s01-in-f18.1e100.net [173.194.38.146]

C:\Users\Vaibhav>tracert www.facebook.com
Tracing route to star.c10r.facebook.com [69.171.224.42]
over a maximum of 30 hops:
1 <1 ms <1 ms <1 ms 192.168.1.1
2 49 ms 37 ms 37 ms 117.223.32.1
3 37 ms 37 ms 38 ms 218.248.173.218
4 140 ms 136 ms 139 ms 115.113.165.93.static-mumbai.vsnl.net.in
[115.113.165.93]
5 137 ms
*
*
172.29.250.33
6
*
*
*
Request timed out.
7
*
260 ms 260 ms 115.114.85.233
8
*
*
*
Request timed out.
9
*
229 ms 229 ms if-2-2.tcore2.WYN-Marseille.as6453.net [80.231.217.2]
10
*
239 ms 230 ms if-9-2.tcore2.L78-London.as6453.net [80.231.200.14]
11
*
230 ms 269 ms if-2-2.tcore1.L78-London.as6453.net [80.231.131.2]
12
*
*
*
Request timed out.
13
*
181 ms 171 ms Vlan533.icore1.LDN-London.as6453.net
[195.219.83.102]
14
*
319 ms 309 ms vl-3604-ve-228.csw2.London1.Level3.net
[4.69.166.157]
15
*
*
*
Request timed out.

16
*
*
*
Request timed out.
17
*
309 ms 309 ms ae-61-61.csw1.NewYork1.Level3.net
[4.69.134.66]
18
*
309 ms 310 ms ae-62-62.ebr2.NewYork1.Level3.net
[4.69.148.33]
19
*
309 ms 309 ms 4.69.135.185
20
*
310 ms 309 ms ae-61-61.csw1.SanJose1.Level3.net
[4.69.153.2]
21
*
357 ms 309 ms ae-1-60.edge2.SanJose3.Level3.net
[4.69.152.17]
22
*
312 ms 311 ms FACEBOOK-IN.edge2.SanJose3.Level3.net
[4.53.210.70]
23
*
321 ms 313 ms ae1.bb01.sjc1.tfbnw.net [74.119.76.23]
24 332 ms 333 ms 332 ms ae16.bb01.prn1.tfbnw.net [31.13.24.254]
25
*
330 ms 329 ms ae0.dr01.prn1.tfbnw.net [74.119.79.101]
26
*
331 ms 330 ms po1022.csw01a.prn1.tfbnw.net [74.119.79.49]
27 330 ms 330 ms 330 ms edge-star-ecmp-01-prn1.facebook.com
[69.171.224.42]
Trace complete.

 In a large scale network, is it possible for a single

PCE
to collect all information of traffic amounts between
all nodes?

The information of traffic amounts between all

nodes is hard to collect in a short interval because

the number of nodes the PCE has to query and
amount of information to be collected are large

Thus, the TE, using the information of traffic

amounts between all nodes is hard to recalculate the

routes against traffic changes that occur in a short
period of time.

How to do it?
One approach is to divide the network into
multiple small ranges
Results in small amount of required traffic
information and reduced calculation time

Can such TE, within the small ranges

handle the large traffic changes?

In networking terms,Traffic is
basically data which flows
throughout various networks
consisting of LANs,WANs etc.
This data is generally in the form of
small units called Packets.

Types of traffic
Constant Bit Rate (CBR)
Variable Bit Rate (VBR)
Bursty Traffic

Data Traffic Descriptors

CBR

VBR

Bursty

Congestion in Network
What is it?
Why is it?
How to avoid it?

Congestion
If the number of packets sent to a
network are more than its capacity to
handle them,congestion is said to
have taken place.
It can occur due to various issues
,most significant being Queues at
routers.
Network Performance is seriously
affected by congestion.

Network performance is measured

with respect to two simple factors:
1)Delay.
2)Throughput.

Packet Delay

Throughput

Congestion Control
Techniques

Open Loop Congestion Control

To prevent congestion before it
happens in any particular network.
It is performed either by source or
destination present within the
network.

Retransmission Policy
Retransmission increases congestion
in network as packet is retransmitted
if no
acknowledgement is received.
Good policy can avoid it.
Proper timer periods can help reduce
congestion.

Window Policy
The type of window selected also
affects congestion.
Selective Repeat window is much
better than Go-Back-N window as
only required packets are resent.
Window Size also has to be selected
appropriately.

Acknowledgement Policy
ACKs increase load on network
Various ACK policies
ACK sent only if there is a data packet to
be sent
ACK sent after special timer goes off
ACK sent for N packets together

Discarding Policy
This policy can be used but with
caution so that it doesnt affect
quality of network.
How many and which packets can be
discarded? should be considered.
Ex: Audio file transmission

Admission Policy
In Various Networks,this policy helps
considerably in controlling
congestion.
Here,the router refuses to establish
new connection if there is a
possibility of congestion within the
network.

Closed Loop Congestion Control

In this method,we try to alleviate
congestion after it happens.
Useful when open-loop methods
seem to fail.

Backpressure

Node to Node congestion control

Travels in the opposite direction of data flow
Only in virtual circuit networks (as we have
to know the upstream router)

Choke Packet

Warning directly to the source (not to

upstream router)
Applied to packet switched networks
Intermediate nodes do not take any action

Implicit Signaling
No communication between source
and congested node
Delay in ACKs --- A sign of congestion
It is guessed by the source which
takes appropriate measures to
reduce congestion.

Explicit Signaling
Congested node communicates with
source or destination
Here,congestion info. is embedded in
data packet,unlike choke packet
method;where separate packet is used
Not like choke packet
Backward Signaling to source
Forward Signaling to destination

Quality of Service

Flow Characteristics

Techniques to improve QoS

Scheduling
Treating all flows coming to router in fair
manner
First-in-First-out Queuing
Priority Queuing
Weighted Fair Queuing

FIFO Queue

Priority Queue

Drawback:Starvation

Weighted Fair Queue

Traffic Shaping
It is a mechanism used to control the
amount and rate of traffic sent to
network.
Two techniques:
1)Leaky bucket.
2)Token bucket.

Leaky Bucket

Leaky Bucket
Implementation

This mechanism shapes bursty traffic to fixed

rate traffic,thus improving QoS.
However,if bucket is full,packets are
dropped.

Token Bucket

Bursty traffic into regulated maximum rate.

Accountability for the free time of source as
well.
Efficient utilisation of empty bandwidth.

THANK YOU.