• Number of servers, location in the network, and
load balancing policies.
• Number of proxies, location in the network, and
cache or splitting policies.
• Transport protocol between each service device
(TCP, UDP, etc).
• Number of clients and location in the network.
• Number of producers and location in the
network (Only in live services with online
generated information)
Resource Definition
In this step the quantity of resources, both in service
devices (servers, proxies, etc) and in the
communication network must be established.
Parameters Definition
In this phase one or more elements must be
determined as parameters. The values for these
parameters are not extracted from the real services,
instead, they will be defined arbitrarily depending on
the goal of the experiment.
Values Definition
In this phase a set of values for each parameter of
the experiment must be defined. The process is, first
to determine the maximum and minimum values and
second to establish the criteria for the intermediate
values. For instance, determining the gap between
them: Linear; Exponential; Free.
When there is more than one parameter, there are
two possibilities to combine them:
• Blind combination: generating all the
possibilities with all the parameters.
• Intelligent combination: eliminating those
combinations that are not interesting.
2.6 Results analysis
The last phase is the results analysis. The results
obtained from lab experiments are not different from
the analysis of the information gathered from the
real service. This analysis is performed using the
tests defined in (Pañeda, 2004).
3 CONCLUSIONS
Lab experiments are very important to predict the
future performance of an internet service. In the case
of audio/video this is even more true because of the
high quantity of resources necessary to provide the
contents with an acceptable quality. Defining a
method to specify the experiments is necessary to
reach accurate results and interesting conclusions.
Nowadays this method is being used to perform
lab evaluations of two real services: LNE TV, the
audio/video service of La Nueva España –which is a
video-on-demand service–, and Asturies.com
Radio, the Internet radio service of Asturies.com –a
live audio service–. Several qualities for the audio
and video contents, service architectures and
quantity of resources are being analyzed. Figure 6
shows the bitrate temporary evolution in the server
output in an experiment for the evaluation of an
Internet radio. Each line represents this evolution for
a different number of users (100 to 700).
0
2000
4000
6000
8000
10000
12000
123456789101112131415161718192021
Bitrate (Kbps)
100 200 300 400 500 600 700
Figure 6: Internet Radio evaluation
ACKNOWLEDGEMENT
This research has been financed by the operator
Telecable de Asturias S.A.U. and the newspaper
La Nueva España within the Media XXI II project,
and the Spanish National Research Program
within INTEGRAMEDIA project (TSI2004-00979).
REFERENCES
Arias, J. R et al. 2002. Evaluation of Video Server
Capacity with Regard to Quality of the Service in
Interactive News-On-Demand Systems. In PROMS-
IDMS2002. LNCS 2515.
Chekasova, L. and Gupta, M. 2004. Analysis of Enterprise
Media Server Workload: Access Patterns, Locality,
Content Evolution and Rates of Change, IEEE/ACM
Transactions on Networking
Pañeda , X. G. et al. 2005. Workload analysis of a video-
on-demand service with a wide range of subjects and
length. In IASTED EUROIMSA.
Pañeda, X. G. et al. 2004. Analysis and Configuration
Methodology for Video-on-Demand Services Based
on Monitoring Information and Prediction. ICEIS.
Jin, S. and Bestavros, A. 2001. GISMO, A Generator of
Internet Streaming Objects and Workloads, ACM
SIGMETRICTS.
ICETE 2005 - MULTIMEDIA SIGNAL PROCESSING
200