Method for Product Lines? Recommendation and Configuration

Thanks to the recent emergence of the cloud computing and services offered by this technology, the economy of the auto-adaptive systems has exploded of several million Euros in 2013. The new insistent challenge that has to be considered in this context is the dynamic adaptation of infrastructure and services based on demand, technical constraints and obstacles such as breakdowns, decreased performance, reduced quality of flows, etc? The project starts from the premise that these services and infrastructure are actually complex and configurable products. The difficulty to maintain an optimal quality of service lies in the choice of the right and optimal configuration among the possible ones. Assuming that cloud services can be considered as dynamic product lines, the project proposes to develop a methodological tool box allowing their specification, their dynamic configuration and monitoring.Recent scientific advances have focused on how to offer customized products (software or other) based on models and algorithms related to what is called "Product Line Management" [BOS00, CN01]. The idea is to allow customers to customize their products by selecting from a list of options depending on their preferences instead of offering them predefined products individually. The customer can juggle with the options, retrace his choice, etc. In order to offer this service, models that specify all the characteristics of a products line, with all options, variants and configuration constraints must be defined. This information is used by "configurators" whose role is to decide whether a configuration chosen by a client as a combination of options, is achievable and how far it respects the Service Level Agreement (SLAs). Even if this function seems to be simple, it is extremely difficult to be satisfied in an industrial level. Indeed, the algorithmic computational complexity grows quadratically (or exponentials in the worst cases) proportionally to the number of options: more options imply more product variants and more difficulties to check constraints [MWC09]. Moreover, in practice, selecting from many options might become annoying for the customer who does not necessarily knows where to start or which right options to choose. Furthermore, when the potential customer sees that his/her choices are not "achievable", ultimately he may turn away to a competitor. Therefore, we need a new approach to design and create a recommendation and configuration (configurable) dynamic system, with a monitoring component. This project proposes the specification of cloud services and infrastructure, described as knowledge ontology for the recommendation and configuration through dynamic interactions. On the one hand, the interest of the recommendation process is to make suggestions to customer by reasoning from known configurations (already realized and/or desirable). On the other hand, the interest of dynamic interaction-based configuration consists in informing the customer, in real time by desirable/possible/inaccessible options according to choices he/she has already made. In this vision, we propose to build various software packages that are adaptable and deployable in the cloud; by offering Dynamic Software Products Lines (DSPLs) [HHPS08]. Moreover, we propose to provide monitoring services in order to guarantee the continual respect of technical constraints, by notifying every violations or perilous situation.The use of DSPL models and ontology raises numerous research questions to solve the online configuration problem: ? What are the necessary knowledge for the recommendation and how to represent it? ? How to link recommendation knowledge (Ontologies) with configuration knowledge (Product Line Model)? ? What learning mechanisms would enhance knowledge about Product Lines? configuration? ? What exactly are the requirements related to the interaction model of online configuration (requirements binding recommendation and dynamic configuration)? ? What algorithms could effectively address the online configuration problem? ? What are the right metrics and techniques to use in order to monitor the software product lines?By answering the earlier issues, the project is expected to contribute to the economic sector of the cloud and online configuration, in many domains such e-marketing, e-commerce, e-government, petroleum industries management, etc... It will combine the research results of the participating teams and the experiences outcomes of the industrial partners in Oman, to produce an innovative approach, based on ontology, for the specification, the recommendation and the monitoring of DSPL which will improve the economic efficiency and income of various industries. To conclude, the project is expected to bring solutions, in terms of services, infrastructure and tools that benefit many Omani organizations and enterprises that are facing the need of software?s customization in the context of online configuration and recommendation and/or the need of services? monitoring and assessment.