A Brief Synthesis of QoS-QoE Methodologies
Hala Rifa¨ ı, Samer Mohammed and Abdelhamid Mellouk
LISSI, EA 3956 - University of Paris-Est Cr´ eteil
122 rue Paul Armangot, 94400 Vitry-Sur-Seine, FRANCE
E-mails: {hala.rifai, samer.mohammed, mellouk}@u-pec.fr
Abstract—This paper presents an overview of some existing
techniques of Quality of Service QoS and Quality of Experience
QoE. QoS stands for the quality of the network in terms of
transporting data with a minimum of delay and packet loss and
a maximum of bandwidth. The extremum values depend surely
of the provided service. QoE defines the quality perceived by the
user and is assessed at the terminal level. It can be computed
through objective and subjective ways. A comparison of the two
techniques as well as an improvement of the service by combining
them are also addressed.
I. I NTRODUCTION
Evaluating a network amounts to the determination of the
quality of the services transmitted through this network from
end-to-end. The services englobe all kinds of data including
the voice, image, video, etc. Some examples of detectable
problems are the freezing of an image, the non-synchronization
of voice and images in video transmissions, the loss of connec-
tion, the delay in communication, etc. The most common way
to achieve the quantification is by attributing a kind of rank for
the transmission going from 1 to 5, 1 reflects a poor service
while 5 reflects an excellent one. This way of measurement is
the so-called Mean Opinion Score MOS.
For many years, the assessment of a network has been per-
formed objectively by measuring and/or computing different
criteria that determine the quality of the network with respect
to the transmitted data (delay, packet loss, bandwidth, etc.).
This quantification is called the Quality of Service QoS of the
network. However, this technique does not consider the user
in the loop. Recently, a new methodology has been introduced
to evaluate a network. It usually takes into consideration the
opinion of the user concerning the afforded services. This
new technique is called Quality of Experience QoE and is
a subjective assessment of the transmission. One should also
note that in the case of multimedia transmission, the evaluation
of QoE is affected by the quality of the data at source, e.g.
a multimedia file, for example, generated with a poor quality
using a non-adaptable compression algorithm can deteriorate
the evaluation.
Recently, many studies have emerged aiming to ameliorate
the QoS of a network based on the user’s perception or QoE
within a closed loop.
The paper is organized as follows. In section II, Quality
of Service mechanisms are detailed. Quality of Experience
methodologies are presented in section III. A comparison of
the two techniques besides of improvements proposed in the
literature are stated in section IV. Finally, some conclusions
are given in section V.
II. QUALITY OF SERVICE
The quality of service QoS consists on defining a set of
technologies called QoS mechanisms that allow the network
manager to ensure an efficient optimization and/or control of
the resources in order to afford the best services with the
lowest cost. It can be determined by the capacity of a network
to convey data in terms of availability, transmission rate, delay,
packet loss, etc. and represents a kind of engagement of the
services providers toward their clients.
Different criteria contribute to evaluate a network QoS. One
can cite the i) delivery guarantee, ii) protocol, iii) latency time,
iv) jitter, v) congestion and vi) bandwidth.
Before developing mechanisms allowing to obtain a good
QoS of a network, let’s first present the traffic classification
problem.
A. Traffic classification
It consists on discriminating traffic into different classes
relative to the protocol used, port number, data type besides
of some statistical information like packets size and arrival
frequency. Examples of classes are: data, multimedia, e-
mail, virus, etc. Each class has its own characteristics going
from attributed priority to flow rate limit, shaping, routing,
etc. These characteristics are, in general, determined by the
network operator. The classification allows to treat classes
differently in order to have a more appropriate management of
the traffic along the network and consequently deliver a better
QoS. For example, traffic classification can be achieved at the
ingress point of a network in order to divide it into elementary
flows, that are managed differently afterwards.
Besides the previously mentioned approaches to achieve
traffic classification, various techniques have been applied.
A widely used technique consists of determining the traf-
fic classes according to network flow parameters obtained
from the packet headers then applying some statistical-based
learning techniques to classify the traffic according to de-
fined classes [1], [2]. Considering that a single level cannot
unveil the generating application, a multilevel approach of
addressing a traffic flow is developed in [3] and is not based
on information based in the payload, therefore its accuracy
is contested. A classification based on cosine similarity of
packets is addressed in [4], hybrid technique in [5] consisting
of determining, within the transport layer, sub-classes within
two main classes discriminated as P2P (Peer-To-Peer) and non-
P2P packets, etc. One should emphasize that most of these
approaches are adaptable to encrypted applications.
978-1-4577-0908-1/11/$26.00 ©2011 IEEE