An efficient algorithm for virtual network function placement and chaining

Oussama Soualah (1, 2, 3) , Marouen Mechtri (1, 2, 3) , Chaima Ghribi (1, 2, 3) , Djamal Zeghlache (4, 1, 2, 3)

1 TSP - RS2M - Département Réseaux et Services Multimédia Mobiles
2 SAMOVAR - Services répartis, Architectures, MOdélisation, Validation, Administration des Réseaux
3 CNRS - Centre National de la Recherche Scientifique
4 R3S-SAMOVAR - Réseaux, Systèmes, Services, Sécurité

The virtualized network functions placement and chaining problem is formulated as a decision tree to reduce significantly the complexity of service function chaining (SFC) in clouds. Each node in the tree corresponds to a virtual resource embedding and each tree branch to the mapping of a client request in some physical candidate. This transforms the placement problem to a decision tree search. We devise a new algorithm based on the Monte Carlo Tree Search (MCTS) to incrementally build and search within the decision tree. Thanks to the proposed SFC-MTCS strategy, an optimized solution is computed in a reasonable time. Extensive simulations assess the performance and show that SFC-MCTS outperforms state of the art strategies in terms of: i) acceptance rate, ii) providers revenue and iii) execution time

Mots clés: Service function chaining; Network function virtualization; Monte-Carlo tree search

Domaines: Réseaux et télécommunications [cs.NI]

Oussama Soualah, Marouen Mechtri, Chaima Ghribi, Djamal Zeghlache. An efficient algorithm for virtual network function placement and chaining. CCNC 2017 : 14th IEEE conference on Consumer Communications & Networking, Jan 2017, Las Vegas, United States. pp.647 - 652, ⟨10.1109/CCNC.2017.7983207⟩. ⟨hal-01629694⟩