A Semi-Markovian Individual Model of Users for P2P Video Streaming Applications

Grégory Bonnet (1) , Ihsan Ullah (2) , Guillaume Doyen (2) , Lionel Fillatre (3) , Dominique Gaïti (2) , Igor V. Nikiforov (3)

1 Equipe MAD - Laboratoire GREYC - UMR6072
2 ERA - Environnement de Réseaux Autonomes
3 LM2S - Laboratoire Modélisation et Sûreté des Systèmes

P2P-based live video streaming has become one of the major applications on the Internet. Due to the importance of the user behavior in these systems, intensive measurement studies have been performed over them in order to provide global models. However, in real networks the individual user behavior does not necessarily follow the global or average behavior. As the measurement studies face difficulties to establish the link between a trace and an individual user due to both technical and privacy issues, we propose individual models of users inspired from fictional characters who represent different kinds of consistent behaviors. In this paper, we define such models with a non-homogenous semi-markovian process and give validation results that show that our model is consistent with global models in terms of population and session duration.

Mots clés: multi-agent simulation; peer-to-peer systems; markovian model

Domaines: Intelligence artificielle [cs.AI]; Modélisation et simulation; Système multi-agents [cs.MA]

Grégory Bonnet, Ihsan Ullah, Guillaume Doyen, Lionel Fillatre, Dominique Gaïti, et al.. A Semi-Markovian Individual Model of Users for P2P Video Streaming Applications. Proceedings of the 4th International Conference on New Technologies, Mobility and Security, Feb 2011, paris, France. pp.1-5. ⟨hal-00952019⟩