Design and evaluation of an autonomic platform for the development of synchronous e-learning activities

1. Introduction

E-learning encompasses a variety of teaching processes and distance-learning techniques which are based on the use of interactive multimedia technologies. In the synchronous modality, a temporal coincidence between instructor and students in the e-learning activities is mandatory, implying interactive communications with strict real-time constraints.

Typically, synchronous e-learning activities have been carried out through multimedia distribution platforms similar to those used for teleconferencing. Nowadays, there is a convergence towards combining both (1) the features of platforms for teleconferencing and online meetings, along with (2) the features of collaborative software and personal organizers. In general, these platforms make efficient use of the available resources and infrastructures. However, several issues related to the complexity of managing and configuring such platforms are still open challenges.

This thesis [7] describes the design of an autonomic multimedia distribution platform for developing synchronous e-learning activities in an efficient way while minimizing the required human intervention. The research behind this thesis has been focused on the autonomic computing paradigm, and its application to the field of real-time interactive multimedia communications. The contribution of this thesis is twofold. First, the complete set of autonomic properties that are found in current MCSs are presented. Then, a taxonomy to classify and compare the techniques to achieve these properties is proposed. Second, a multimedia distribution platform including the most critical autonomic properties is designed and implemented. The platform uses various heuristics to cope with optimization and healing challenges. The quality of experience observed by users while using the platform has also been assessed in a real synchronous e-learning activity.

2. Autonomic computing

The topic of Autonomic Computing (AC) was presented by Horn to alert the scientific community and the IT industry to the increasing complexity of computing systems and software, and the looming crisis that this would entail [4]. The main purpose of an AC system is self-management. In the early stages, autonomic systems were described according to a set of classic characteristics usually referred to as self-properties: self-configuration, self-optimization, self-healing and self-protection [5]. From the beginning, nevertheless, it was predicted that such a set of self-properties would increase [10].

New self-properties are proposed as research into autonomic computing progresses, increasing the initial set of four self-properties. However, this thesis points out that the wide range of new self-properties conceptually differs from the initial self-properties. The new self-properties can be considered as pillars to reach one or several initial self-properties. In this thesis they are referred to as self-subproperties. Thus, a self-subproperty can be understood as an autonomic behavior or capability from which a self-property emerges.

The most prominent MCSs [1,6,9,11], among others, have been surveyed to identify common self-subproperties [8]. The contribution of this research has been threefold. First, various autonomic self-subproperties and the relations between them and the former self-properties have been discovered. Second, the implementations of these self-subproperties in the MCSs have been analyzed. Finally, a taxonomy has been proposed to classify the wide variety of techniques for implementing the discovered autonomic self-subproperties.

3. Autonomic platform

Various self-management capabilities have been included in a multimedia distribution platform related to self-healing and self-optimization. Concretely, the platform has been designed including techniques to achieve several self-subproperties: self-organization, self-diagnosis, self-stabilization, self-deployment and self-regulation. The platform automatically deploys a mesh of reflectors to forward data between members. Each reflector is located in a multicast island, so IP multicast can be used to deliver data from reflectors to members. Whenever a reflector fails, the members served by the reflector must be re-directed to other reflectors. A heuristic chooses the reflector to which a member has to be re-directed by estimating the resources available in each reflector and the impact on the overall performance. When a new reflector joins the mesh of reflectors another heuristic is used to optimize the data delivery. As a result, some members may be re-directed to this new reflector to decrease the resources consumed in the rest of the reflectors. The efficiency and resilience of the autonomic platform has been exhaustively tested using the NS3 simulator [2].

Furthermore, the autonomic platform has been deployed to support the training of human resources in ArcelorMittal Spain jointly with a synchronous e-learning tool. The users agree on the suitability of the platform to convey synchronous activities and report a successful user experience [3].

4. Conclusions

Most the prominent MCSs include autonomic capabilities. The analysis of the techniques to implement these capabilities shows that self-organization is by far the most common in MCSs. The autonomic platform developed includes additional autonomic capabilities, which provide an efficient and resilient data delivery service without human intervention. This has been proved with extensive simulations and user tests.