TMA portal

Orange Labs Technical Coordinator: Sergio Beker

Technical coordinator email address: sergio [dot] beker <at> orange-ftgroup [dot] com

Site: Sophia Antipolis, France

Doctoral Thesis Subject (Title): Modeling Quality of Experience measuring functions for service composition SLA.

Global context and state of the art

Service Level Agreements (SLAs) have historically been defined in terms of technical service implementation parameters. In the classical sense, an SLA includes a list of parameters related to the network performance and in which the objectives are defined as individual thresholds on those parameters. In the context of business services market, the evolution towards on-line service platforms (i.e. in SaaS, ASP or application outsourcing modes) and the associated composite service offerings makes the SLA hard to define and implement. The diversity of services and actors is one of the main causes of such complexity. Currently, the SLAs are tailored to each customer, often composing vertical service levels on metrics related to the technical service implementation rather than the global user perception or the service utility. Those solutions, besides of being costly for the service provider, do not correspond to the current market needs. Economies of scale are made difficult because of the low degree of reuse of the different service definitions. The current solutions on the service level management (SLM) domain lack of tools and methodologies on the elaboration phase of the SLA life cycle.

A number of studies have however shown that meeting the individual service levels on the service performance metrics does not guarantee the final user satisfaction. The user perceived quality depends not only on the set of performance parameters, but also on the human factors related to the usage context of the service. The SLAs must then evolve to take into account those metrics for Quality of Experience (QoE).

A methodology for SLA creation, allowing for identifying the service components and attaching the corresponding QoS performance metrics has been developed within the DIAM laboratory. A QoE Framework, issued from the same research efforts, allows for a homogeneous usage of QoS and QoE terminology, identifying reference interfaces, actors and roles of the intended SLAs. However, the expressions for QoS and QoE as a function of network performance metrics are only available for a small set of component services (e.g. mostly related to voice or video services). On the other hand, the SLA elaboration methodology for composite services evolves in a multitenant and hybrid environment (i.e. services are composed following a business process view: components are not all provided by the same actor and are of different nature).

Thesis Objectives/Expected Results/Scientifical and Technical Challenges

Models for QoS and QoE for composite services and network services in general are a difficult scientific challenge that needs to be addressed. Analytically determining measuring functions translating QoS into QoE is extremely complex due to the number and the different nature of intervening parameters, as well as their interdependence. The subjective tests methodologies could greatly benefit from service models capable of taking into account the more subjective factors of quality appreciation alongside the more technical ones. This thesis aims at:

• characterizing the different applications according to service types/categories (for example those proposed by ITU-T),
• validating and proposing evolutions to existing models for QoS and QoE for the service types/categories (e.g. web services, interactive applications, etc.),
• validating and proposing evolutions to existing metric composition rules for QoE,
• validate and propose evolutions for existing subjective test methodologies allowing to unveil the dependencies among the different human factors from the service models and the user quality appreciation in a more systematic way.
• Depending on the results obtained for the service models and the corresponding QoE estimators, the PhD candidate would be encouraged to develop a testbed for automating the process of obtaining the QoE estimators for a given application from the corresponding service model. Obtaining the parameters characterizing the estimator would also need the definition of the corresponding subjective tests, as well as the tools necessary to analyze the results.

Methodological approach proposed by the technical coordinator

In a context of service composition, the first objective is to obtain individual models for QoS and QoE in terms of performance parameters for the different service categories. Then the measuring functions for the QoS and QoE for the composite service would be obtained from the metrics associated to the component services through composition rules.

The doctoral thesis is in synergy with the current work within the RDU SPO around the perceived quality through the contribution to different projects about evolved SLA elaboration for composite services and the corresponding QoS/QoE. In particular, a QoE framework and a composition methodology are being defined from existing results and current standardization efforts (ITU-T, ETSI) which take into account the human factors in modelling the perceived quality. This general framework aims at obtaining homogeneous definitions of QoS and QoE, situating the different models relative to the different service categories. It is mainly oriented to setting a reference for the different works on the domain (QoS, QoE, anomaly detection, etc.). A normalization effort around the framework needs to be discussed, depending on the group strategy and the current efforts within the RO "QoS in the Network".

The candidate for this doctoral thesis is expected to have good knowledge on the following topics:

• Performance evaluation and network dimensioning, queuing theory
• Statistics, probabilities
• Basic knowledge on network protocols and architectures
• Model engineering (MDA, PIM PSM, UML)
• A user oriented vision (human factors) is a plus.

Global Planning

The doctoral thesis will be mainly structured as follows:

• A bibliographic study detailing the methods available on the research community and standardization bodies, relevant to the work currently being carried out within the RO. This study would constitute a baseline for the proposition and benchmarking of measuring functions for QoS and QoE for the different service categories.
• Conducting simulations and subjective tests in order to make appear the impact of the different intervening parameters. This phase would be conducted in collaboration with other projects within the RO more oriented to the user modelling and the subjective aspects of quality.
• Study the relationship between the QoS and QoE indicators from the data obtained through the experiments. The correlations might be studied via different analytical and statistical methods. The target is to propose models for the measuring functions of QoS and QoE for the different studied service categories, and to validate the theoretical models obtained through the composition methodology we have developed.
• Propose and validate the necessary evolutions to the metric composition methodology.
• Plan and develop (at least specify) a plat-form implementing the service models and characterization tools for the service types/categories studied.

  Other Contributions (contribution to collaborative projects if any)

• Contribution to the WebQoE Project currently active around the modelling of QoS -> QoE measuring functions for Web Services.
• Participation to COST actions (SIG QoE within the COST-TMA action and a new QoE COS action: QUALINET recently submitted).
• Contribution to future projects to be defined.