Computer networks-ATM Traffic Management

2.  Traffic Management is the act of managing network
traffic , providing service guarantee to user
connections and ensuring optimal utilization of
network resources

3. .
• Traffic Contract Management.
• Traffic Shaping
.
• Traffic Policing
• Priority Control
.
• Flow Control
• Congestion Control

4. Traffic Contract Mng.
Traffic Shaping
Traffic Policing
• Connection admission
control(CAC)
• Leaky bucket
• Usage passage control
• Network passage control
Priority Control
Flow Control
Connection Control
• Selective cell discard
• Generic flow control
• ABR flow control
• Frame Discard
• EFCI

5.  It provides the means to allocate network resources
to contending user connection and to manage
network resources efficiently (i.e. deciding whether a
new connection request can be accepted or not).
 It is defined by two set of parameter : Network traffic
descriptor and QoS Parameters.
 It uses “Connection Admission Control”
mechanism.
 Set of actions taken by the network at the call
establishment phase in order to ensure whether a
VPC or VCC to be accepted or rejected.

6.  Decide whether new user call request is to be
accepted or rejected.
› CAC ensures sufficient resource available in the network to
support the new requested call
› CAC also ensure performance of existing connection should
not degrade by accepting new call.
 If sufficient resources are not present or acceptance
of connections may endanger QoS of existing
connections , network rejects the call.
 CAC also ensures resources are not allocated in
excess to existing capacity.
 Without CAC algorithm, unchecked admission lead to
Congestion(where demand>availability).So, CAC act
as preventive congestion control mechanism.

7. Non Statistical Or Peak Bandwidth Allocation
Algorithm: it allocate resources at the peak cell rate of the
connection.
 Merit : It is simple to implement
 Demerit : lead to gross underutilization(CAC will allocate
resources to peak rate but data which will not arrive at peak
rate suffers).
Statistical CAC Algorithm : Using statistical allocation the
sum of peak rate of all the connections can exceed the link
capacity. A parameter called Statistical gain measure
advantage over non statistical algorithm (i.e.) ratio of
connection admitted with statistical multiplexing to the
connection admitted with non statistical allocation.
 Merit : These connections do better resource utilization &
performance efficient.
 Demerit : Difficult to implement.

8.  Minimum processing complexity: Because ATM switch
operates in real time and require connection establishment
and release requests to be served with minimum latency.
CAC algorithms should be easy to implement.
 Model independence : CAC algorithm should be model
independent Because as we have seen 2nd element of
traffic management is traffic shaping , in altering shape of
model , CAC algorithm will also Alter.
 Balance between cell loss and link utilization: cell loss
directly concern with user and link utilization concern with
network. Low link utilization and greater bandwidth
allocation lead to low cell loss or vice versa.
 Algorithm should be scalable(computation required should
be independent of number of connections).

9.  It provides the means to alter traffic characteristics
of connection and deliver a more predictable,
conforming traffic to the network.
 It uses “Leaky Bucket” mechanism.
 Mechanism that alter the traffic characteristics of
stream of cells on VCC and VPC to achieve
desired traffic characteristics , in order to ensure
conformance or to achieve better network
efficiency to meet QoS objectives

10.  (HOST COMPUTER)
Bursty Flow(PACKETS)



Discarded
drop of water
Fixed flow of water
 Leaky bucket shapes Busty
traffic into fixed rate traffic by
averaging data rate.
 In this mechanism peak cell
rate is reduced to get
Stability.
 It makes traffic characteristics
more predictable.
Bucket
Tap
Source
system
Gives busty
data
Data stored
in buffer
Lowers the
data rate
Stabilize
traffic with
lower peak
cell rate

11. ARRIVAL FUL
L?
Y
E
S
NO
QUEUE
LEAK
Y
PROCESSOR
0 2 7 10 0 10
12
mbps
Time:
0-2
12*2:24
mbits
2
mbps
2*3:6
mbits
After algo
3 mbps
Time :0-10
3*10:30mbits

12.  It provides means to ensure that the connection
adheres(compatible) to the parameters of its traffic
contract and non conforming connections are
penalized.
 It uses “usage parameter control/network
parameter control” policing mechanism.
 Set of actions taken by the network to monitor and
ensure that the traffic contract is represented in
terms of traffic offered at the user access and the
network access respectively.

13.  Mechanism to monitor traffic and enforce the traffic
contract is called as usage passage control(UPC).
 Depending on traffic monitoring at UNI called UPC or
NNI the monitoring functions called NPC respectively.
 UPC applies to both user connections and signalling
channels, implies along with controlling amount of
user cell ,it also monitor amount of signalling
information entering the network.
 Commonly used algorithm is GENERIC CELL RATE
ALGORITHM(GCRA).

14.  It define conformance of cell with respect to negotiated
traffic control(conformance means cells arrive either at
time or after).
 Defined in terms of 2 parameters i.e. GCRA(I,L):
› Increment(I) that specify when next cell is excepted.
› Limit (L) that specify permissible deviation from expected
arrival time.
 Checks conformance of traffic contract using twin
concept i.e.
› Expected arrival time =Theoretical arrival time (TAT).
› Actual arrival time.
 All cells arriving before time are marked as non
conforming cells

15. Conformance of a cell is decided on the basis of TAT as following:
1st incoming cell initialize algorithm
Set TAT of cell 1 .......(.i.e.) TAT1 = ta(1).
For kth cell conformance is decided on basis of TATk,
CASE I if k arrives after TATk cell is declared as conforming.
For k+1 cell TAT = TATk+1 is set to ta(k)+I , k is
incremented.
CASE II If cell arrive between TATk and TATk-L cell is
declared as conforming.
TAT for next cell will be TATk+I not ta(k)+I.
CASE III If cell arrive before TATk – L  cell is declared as Non-
conforming.
TAT for next cell is TATk.

17.  Bucket of capacity (I+L), where each cell pour I data
into it.
 Bucket is leaking @ 1data unit/time unit.
 Capacity of bucket is X at any time instance.
 Incoming cell is conforming if at its arrival the fluid in
bucket <= L, otherwise it is non conforming.
 Algorithm maintains last conformance time(LCT)
which store the last time the conforming cell was
accepted.
 The amount of fluid drained since LCT is subtracted
from bucket capacity i.e. X*=X-(ta(k)-LCT).

18.  Cell passing – cell simply pass through, network
is lightly loaded . It result in better utilization of
network resources, but while overloading it is not
preferred.
 Cell tagging – cell loss priority (CLP) bit in ATM:
CLP=0 (high priority/conforming cells) CLP=1
(low priority/non conforming cells).Marking
of CLP bit in ATM header marked as 1.
 Cell discarding – simply cell dumping.

19.  It ensures that the Packets with higher
priority should be preferable over the
Packets with lower priority.
 Processing every cell according to the
priority will mean additional delays ,
which might not be acceptable to network
users.
 It uses “Selective Cell Discard”
mechanism.
 Mainly this algorithm is based on
CLP(Cell Loss Priority) bit of ATM header
, as we know ATM Header include 48 bits
payload and 5 bits header(1bit for
Header error control=CLP).

20.  Mechanism discard cell based on priority .
 Cell with CLP bit = 1, consider as low priority.
 Cell with CLP bit = 0 , consider as high priority.
 If network is forced to drop some cells , the network
selectively drops the cells with lower priority.
 Example: layered video encoder ,having unequal
priority, where two data streams are produced having
essential(provide average picture quality) and
enhancement layer(improve picture quality).
• 0011
• 1100
1
• 1111
• 0000
0
• 1010
• 0101
1
11110
00000

21.  It ensure the synchronisation between sender and
receiver i.e. sender’s sending capacity should not be
greater than receiver’s receiving capacity.
 No hop by hop flow control for error correction.
 Used to optimize bandwidth utilization by feed back
mechanism.
 Feedback mechanisms are “Generic Flow Control”
and “ABR Flow Control” mechanism.
 GFC provides a link level flow control mechanism to
control data offered at UNI.
 ABR flow control is a fast resource management
mechanism to utilize any lower bandwidth in the
network for connection belonging to ABR service
category.

22.  It is a link level flow control scheme to control amount of
data flowing at the UNI.
 Applicable to data flowing in forward direction (user to
Network).
 In UNI cell header , there are 4 bits for GFC across UNI
and access control with in customer premises
network(CNP).
 GFC used to control flow of traffic from various ATM
connections and to limit the transmission capacity at the
UNI interface.
 Scope of GFC depend on configuration of end system .
It can be controlled configuration or uncontrolled
configuration.
 GFC is not for uncontrolled system equipment.

23.  Controlled connection: connection whose cell
emission rate is controlled by GFC signals . No
resource is reserved in advance for controlled
connections. Connection belonging to ABR,GFR
etc services are examples of connections that can
controlled.
 Uncontrolled connections: connections whose
cell emission rate is not controlled by GFC signals
(except HALT), resources are reserved in
advance. Connection belonging to CBR service
are generally uncontrolled.

24. • Terminal is
controlled and
controlled ATM
connection
Group B
• Terminal is
controlled and
controlled ATM
connection
Group A
• Terminal is
controlled . Cell
is assigned on
an uncontrolled
ATM connection
• Terminal is
uncontrolled .
Cell is assigned
on uncontrolled
ATM
Connection.
0000 0001
00110101
Bit values for GFC bits:

25.  GFC procedures assumes master slave relationship
between network and user, where network referred as
controlling and end system as controlled equipment.
 GFC signals are as follows:
› HALT: To stop transmission of assigned ATM layer cells for
both controlled and uncontrolled ATM connections . This
signal is sent to limit the amount of cell generated by ATM
end system at UNI. E.g.. HALT and NO_HALT signal is sent
alternatively , transmission capacity reduced to half.
› SET: controlling devices issues SET signal to counter to
specified integer value . Counter specifies the maximum
data the controlled device can send.
› NULL: control devices send this to show no action is taken.

26. GFC Bits in network to User Connection
1 2 3 4
Indicates HALT(1)/NO_HALT(0)
IN 1-q model , set to 0
IN 2-q model , indicates SET(1)/NULL(0)
for Class B
In I-q model, indicates SET(1)/NULL(0)
In 2-q model , indicates SET(1)/NULL for Class A
Reserve
for future
use , set
to 0

27.  ABR stands for “Available bit rate”.
 ABR does not provide any QoS guarantee but It ensures
efficient means to access bandwidth available in network
for which a flow control feedback loop is used.
 Two control approaches are there:
› Credit based flow control: HOP BY HOP flow control
mechanism. Source receives credit from next hop which specify
the amount of data to be sent by source. For every cell sent ,
sender decrement the credit available. At credit=0,sender
prohibited to send next data.
› Rate based flow control: END TO END feedback flow control
mechanism. There is one feedback loop from the source to
destination. Destination send control information to source and
intermediate node can change it and forward it to the source.

28. Credit-based
Desirable because in
temporary condition ,
all nodes have to stop
sending credits to
previous hops.
Require per – link per
VC monitoring , which
complicates the
design and hampers
scalability.
Rate-based
There will be delay in
feedback from
destination to source
even in case of
temporary condition.
Simple , Scalable and
does not require state
information of a no. of
VCs to be stored in
intermediate switches.

29.  “CAC” is an example of preventive mechanism.
 It provides means to ensure the total traffic on a
network is below its carrying capacity.
 It uses “Selective Cell Discard”, “ Frame Discard”
and “Explicit Forward Congestion indication”
mechanisms .
 Selective Cell Discard based on CLP(Cell loss
priority).
 Frame discard policy discard full IP packet instead of
a cell.
 Explicit forward congestion notification used to assist
the network in avoidance of and recovery from
congestion state.

30.  Large IP packets are segmented into number of cells.
 During congestion if one of the cell belonging to IP
packet is dropped, the remaining cells carried by ATM
network become dead cells because they don’t
reassemble properly, this can cause congestion to great
extent.
 So to discard a frame it will be enough to discard 1 from
it.
 Now to tackle this problem of frame discard few
schemes are there:
› Partial packet discard
› Early packet discard
› Random early discard

31. WORST CASE BEST CASE

32.  Partial packet discard(PPD): if one cell from a frame
is discarded whole frame is dropped than. Last frame
cell is not discarded as it indicate loss to the receiver.
› Drawback: Some of the dead cells are still carried till the
destination .
 Early packet discard(EPD): an entire frame is
discarded instead of carrying dead cell to the receiver
and avoid overflowing of the buffer .
› Drawback: lead to low bandwidth flow.
 Random Early discard(RED):Per flow buffer
accounting .It achieves fairness along with early
packet discard benefits. Whenever the queue size in
buffer exceeds certain threshold ,incoming packet is
dropped with certain probability.

34. Congestion Control Flow Control
Global Issue. .
Joint responsibility
of user and network
 Main task is to .
ensure that subnet is
able to carry the
offered load
Local Issue .
Issue between .
sender and reciever.
Main task is to .
ensure that a fast
sender does not
overwhelm a slow
receiver.

35.  Congestion notification is sent to the destination.
 In this congestion bit in ATM header is set to 1 to
indicate congestion in the network.
 After receiving this notification destination
communicate to source to slow down the
transmission rate (in source to destination
direction).
 This depend upon the protocols running at the end
system how destination conveys this information ,
no standard procedure exist .