Project Topics

www.seminarsonly.com

Engineering Projects

Published on Nov 30, 2023

Introduction

In this paper, a new congestion control technique for ATM networks is presented. The technique includes admission control, and traffic shaping. The network traffic consists of real-time traffic and data traffic. Call acceptance is based upon the effective bandwidth and data traffic flow is controlled by effective buffer. Effective bandwidth for a switching node is defined as a vector of bandwidth and an estimated maximum delay at the node. Effective buffer is defined as a scalar of buffer size. The proposed scheme is analyzed by simulation and the results are presented in comparison with other studies under similar traffic conditions.

Proposed Technique

The proposed congestion control technique features parameterized call acceptance. Available bandwidth and maximum node delay are two crucial parameters used for setting up connections. Bandwidth is pre-allocated for real-time traffic based on prescribed mean bit rates. Available buffers are the control parameter for admitting non-real-time cell transfers on a link-by-link basis. Details are shown below [4].

1. Two types of traffic are defined in the model: a) Real-time Traffic (RT): Cells of this type are delay-sensitive. They must be delivered to the destination within a predefined time frame. b) Data Traffic (DT): Cells of this type are delay-insensitive, but they are loss-sensitive. All cells must be delivered.

2. EB ( E ffective B andwidth) is the criterion used for call acceptance. There exists a separate EB for each type of traffic and for each node. EB is a two-element vector with the format of EB = (x, y).

The EB of a node is defined as follows:

EB i = (C AVAIL i , M i ) (1)

Where

EB i = the EB of node i

C AVAIL i = the available (unallocated) channel capacity at node i

M i = the maximum node delay at node i

Note: For simplicity, all definitions in this model are time-implicit. The time factor is syntactically omitted but intuitively understood. For example, EB i is short for EB i ( t ), denoting the EB of node i at time t (the time node i is inquired).

The EB of a RT traffic is defined as:

EB RT i,j = (B RT i , D RT i,j ) (2)

Where

EB RT i,j = the EB of the i th RT traffic at node j

B RT i = the prespecified mean bit rate of the i th RT traffic

D RT i,j = the allowable maximum node delay of the i th RT traffic at node j , and

D RT i,j = D RT i,pred(i,j) - M pred(i,j)

Where

pred(i,j) = the predecesor of the j th node of the i th traffic

The EB of a DT traffic is defined as:

EB DT i,j = (0 DT i , D + DT i,j ) (3)

Where

EB DT i,j = the EB of the i th DT traffic at node j

0 DT i = the prescribed mean bit rate of the i th RT traffic is zero (at the connection

setup time)

D + DT i,j = a quantity that is larger than the allowable maximum node delay for the

i th RT traffic at node j

EB 1 = ( x 1 , y 1 ) and EB 2 = ( x 2 , y 2 )

A RT connection request is granted only if its EB can be satisfied by all intermediate nodes on the route; i.e., RT i can be granted its connection request only if O(EB j , EB RT i,j ) = 1 is true for all j 's on the routes. A DT traffic is also connection-oriented. However, a DT connection request is always granted. From the EB definition for DT traffic (Definition 3) we know that acceptance is instantaneous. In this case, a route can be selected randomly by the entrance node. EF ( EF ective buffers) is the major criterion used to grant cell transfer requests for DT traffic from node to node. There exists a separate EF for each DT cell transfer request and for each node. EF is a scalar quantity.

The performance of the proposed congestion control technique is evaluated by using simulations. We assume the following for all our simulations: a) Channel capacity allocation is based on the prescribed mean arrival rate for each input source. b) RT traffic is shaped by employing a leaky bucket method [5] based on the channel capacity allocated. In our simulations, the leaking rate of a leaky bucket queue coincides with the service rate for that queue. c) Each DT input source is allocated a large buffer (a fat bucket policy) to accommodate sudden bursts of cells without risking any loss. d) The system is in equilibrium and running at full speed (all channel capacity allocated) when it is analyzed