Side_1_360

The DAG monitors are located such that every

packet sent and received by the GenSyn traffic

generators are captured. The location of moni-

tors is shown in Figure 15. This instrumentation

enables accurate measurements to be taken of

necessary performance metrics like throughput,

unidirectional delay and loss.

3.3 Test Scenario Example

In order to describe the main deployment of

GenSyn, this section will describe a complete

scenario where the objective is to test the QoS

of various service classes under different QoS

support strategies. The focus is on the end-to-

end performance, particularly for the real-time

classes. Several traffic scenarios should be

defined for testing of the QoS mechanisms

in an IP network. In this section the different

components of the scenarios are described:

• Type of application (Web, FTP, VoIP, MPEG,
  CBR, etc.);


• Network configuration (best effort, differenti-
  ation);


• Protocol mixture (UDP/TCP ratio);


• Load level;


• Routing (balanced or single bottleneck).

3.3.1 The GenSyn Application Models

The following models are currently available in

the GenSyn framework.

TCP traffic– adjusts to the network perfor-

mance (slow-start window mechanism)

• Web– a model of users (clients) that down-
  load web-pages with all their content (inclu-
  sive applets and images) from real web
  servers all over the world. The url addresses
  are found in a parameter list of predefined
  addresses that may dynamically be updated
  as the experiment evolves.
  
  
  • FTP– a model of users (clients) that down-
    load real files from a server. The files are
    specified in a parameter list of files.
  
  
  
  

UDP traffic– no network performance adapta-

tion at transport level

• VoIP– a model of the information/media
  stream from VoIP users. It sends a determin-
  istic stream of packets (fixed size and inter
  packet arrival time) from each of the active
  users. The model does not include the call
  set-up and disconnection phases.


• MPEG– a model of a video server that is
  sending MPEG-1 coded video sequences
  [Rose95]. All video frames are converted to a
  number of fixed sized IP packets sent back to
  back. The interframe distance is a parameter
  with a default value as recommended by the
  MPEG-1 codex standard. The clients are
  implicitly modelled only as incoming requests.


• CBR– a model of a multiplex of deterministic
  streams of packets with phase shifts.

New models can be defined on request, and the

current models can easily be changed if this is

requested. In Figure 18 snapshots from the Gen-

Syn visualizer are included. The figure illus-

trates that GenSyn can generate packet flows

with very different traffic characteristics.

3.3.2 QoS Mechanisms and Service

Differentiation

There is a trend towards building communica-

tion platforms for integration of a great variety

of applications with different traffic characteris-

tics and users with different Quality of Service

requirements. There is a trend in networking

towards Full Service Network, i.e. a network for

all types of services. The various services and

applications need to be treated differently, and

hence some means for service differentiation is

required with different support for traffic man-

agement and control. In an IP based network

such differentiation and management techniques

are still a research topic. The proposed, and

a) FTP trace b) VoIP trace

c) CBR trace d) MPEG trace

Figure 18 Snapshot of 4
traces captured from the
visualizer in GenSyn