We are happy to report that a PRISTINE paper lead by the Telecom SudParis team has been accepted at the IEEE AINA 2017 conference! The paper, entitled “Hierarchical Small World Overlay for Efficient Forwarding in Volunteer Clouds” focuses on optimizing the forwarding process and managing the connectivity in a scalable and dynamic fashion in volunteer clouds.
Volunteer clouds is a new paradigm of cloud computing where volunteer resources, i.e. user machines are used instead of (or in addition to) dedicated resources, i.e. data centers. Although volunteer clouds offer potential benefits in terms of energy and cost saving as well as elasticity, the volatility and the scalability of the volunteer resources are still open challenges that need to be studied. In this paper, we focus on optimizing the forwarding process and managing the connectivity in a scalable and dynamic fashion in volunteer clouds. We show that volunteer clouds could be modelled by a small-world structure, characterized by short average path lengths among nodes, and strong local clustering, which implies small latency between nodes and therefore robust forwarding in the cloud. First, we propose to build a hierarchical small world overlay network supporting scalability and dynamicity constraints of volunteer clouds. Second, we explore the benefits of the use of the Recursive InterNetwork Architecture (RINA) as a networking solution. The advantages that RINA bring comparing to the IP protocol is its recursiveness and its layer management functions. RINA is based on a ”divide and conquer” strategy enabling scalability over large networks. On the other hand, in RINA there is a common layer management protocol that could be leveraged to design a flexible and dynamic forwarding management layer to maintain our hierarchical small world overlay network. Finally, simulation experiments are provided to evaluate the performance of our proposal. The results show that our small world overlay network gives better performance by reducing the routing table size and the end-to-end latency compared to a random based overlay topology.