Journal Information
  Resources

 

Smart Workload Automation by Swarm Intelligence within the Wide Cloud Computing

Samia Chehbi-Gamoura

Faculty of Economics and Management, University of Lumière Lyon 2, Bron 69500, France


Abstract: The workload automation supported by cloudified environments begins to be agreed by the industry; however there are numerous prevailing challenges which have not been fully addressed yet. The primary of these issues is load balancing that is mandatory to equilibrate workload of resources and to avoid services unavailability. This paper presents a new model, called SWAAN (Smart Workload Automation by Ant-inspired Networks) with a novel algorithm that is capable to ensure the failover. This search algorithm, titled as PB-DNA (Propagation and Back-propagation Diffusion through Neighborhoods Algorithm) can play a role in reaching reliability of the wide cloud environment by considering hot migration of failed tasks in external load balancing (inter-clusters). PB-DNA belongs to the ACAs (Ant Colony Algorithms) that are inspired by the behavior of some species of red fire ants in situations of threat. Unlike stigmergic approaches, the proposed heuristic is accomplished by direct communication diffusion of ant-agents software threads. A deep description of the proposed PB-DNA algorithm is provided with key concepts of ant-agents settings, neighborhood, and tree of local minima. Furthermore, this work offers an application frame of a real case study extracted from IT infrastructure of the French railway company SNCF®. The makespan and downtime benchmarks are measured for PB-DNA and compared with RRA (Round Robin Algorithm) and RSA (Random Scheduling Algorithm). The obtained results place PB-DNA as the most robust approach that produces less unavailability.

Key words: Wide cloud computing, workload automation, SWAAN (smart workload automation by ant-inspired networks), PB-DNA (propagation and back-propagation diffusion through neighborhoods algorithm), swarm intelligence.
Download: Purchase PDF - $ 5