04-10-2012, 03:50 PM
TCP TRAFFIC SHAPING in ATM NETWORKS USING GCRA ALGORITHM
TCP TRAFFIC SHAPING in ATM NETWORKS USING GCRA ALGORITHM .docx (Size: 79.02 KB / Downloads: 24)
Abstract
Simulation results concerning the performance of the TCP protocol implemented in high-speed ATM networks are presented. A simple ATM network is concerned, comprising four TCP connections share the bandwidth of a bottleneck link with some background traffic. In our simulation scenario, the background traffic level and node buffer size are taken as a variable parameter. Both the background traffic and the TCP traffic are shaped according to the GCRA algorithm. The effect of the background traffic on the TCP protocol performance is studied, varying for node buffer size and background load.
Introduction
In the future, the data traffic is expected to be a relevant part of the load in ATM networks. That is, the TCP will be the most frequently used transport protocol in the ATM environment . Our work concentrates on the effect that the heterogeneous traffic present in the network, that we call background traffic, may have on the TCP performance. The importance of the presence of background traffic goes beyond the reduction of the bandwidth available to TCP, since background traffic interferes with the TCP behaviour by altering the probability of cell losses within node buffers. The results presented in this paper are obtained via simulation, implementing a TCP layer within an ATM network simulator named CLASS (Connection Less ATM Services Simulator).
Traffic Shaping
The user of the services offered by the ATM network may perform some kind of control on the traffic it generates in order to verify that it sticks to the negotiated parameters and will thus not be rejected by the network. This operation is clearly traffic shaping, because the operation performed is aimed at the modification of the traffic very similar to the traffic policing, but it is preventive and not repressive and is generally called characteristics, generally by smoothing its burstiness, i.e.to give a pre-defined shape to its profile. Traffic shaping has three main purposes.
These are:
1. The characteristics of a flow of cells that has endured a shaping process are much easier to be described, thus the shaping of the traffic allow the user to negotiate the transmission parameters more easily.
2. If the shaping algorithm is known to the network too, it helps also the network in the management of the call acceptance because it can more easily predict the behavior of the generator.
3. Monitoring of the traffic on the network side, i.e. traffic policing, is much easier and reliable if the input flow has known characteristics.
A number of shaping policies and algorithms have been proposed in literature. But, each one have some advantages and drawbacks. The traffic control algorithm that has been chosen for the inclusion in this study is the Virtual Scheduling Algorithm, This algorithm is basically a token bucket . In any case, although the basic algorithm is the same the actions that are taken by a shaper are radically different from those taken from a policer (``bad'' cells are delayed and not discarded). Each time a cell becomes available for transmission, the Virtual Scheduling Algorithm determines whether the cell is conforming with the Traffic Contract of the connection. The VSA not only provide to control the traffic characteristics, but also provide a means for the formal definition of traffic conformance to the Traffic Contract. The VSA requires only the definition of two parameters: the time increment between cells T and the time limit. The time increment between cells is clearly the time that should pass between two consecutive cells if the traffic was generated at onstant bit rate. τ is the time jitter that is allowed and is clearly related to the parameter generally called ``cell delay variation''. The VSA keeps track of a Theoretical Arrival Time (TAT ), which is the ``nominal'' arrival time of the next cell assuming equally spaced cells hen the source is active. If the actual arrival time of acell is not ``too'' early relative to the TAT , in particular if the actual arrival time is after TAT time of the first cell be ta (1). In this case the cell is transmitted immediately and the TAT is initialized to the ta (1) + T . For all subsequent cells one of the following three alternatives is given: 262 1-4244-1457-1/10/$25.00 © IEEE This full text paper was reviewed at the direction of IEEE Communications Society subject matter experts for publication in the IEEE CCNC 2010 proceedings.