Mathematical Optimization Methods for Research Context NFV and - PowerPoint PPT Presentation

Contributions to Cloud and NFV Areas Makhlouf Hadji Mathematical Optimization Methods for Research Context NFV and Cloud Resource Allocation VMs Placement & Problems Repacking Our contribution Conclusion and Further Research HDR Defense Critical Resource by: Allocation Our Contribution Makhlouf Hadji Conclusion and Future Work VNF-FG Placement Main Contributions 3 March, 2017 Conclusion Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 1 / 69

Research Activities, Projects and Collaborations Contributions Collaborations : with more than 30 researchers to Cloud and NFV Areas Makhlouf Involved in 8 R & D projects (Orange, Nokia,...) More than 50 papers Hadji Mentoring 3 Patents filed Research Context of 6 PhD VMs students Placement & Repacking to.. 2005 2006 2007 2008 2009 2010 2011 2012 2013 2017 Our contribution DEA Conclusion and Further Research Operations Research Research PhD thesis Head of Critical Engineer Researcher Resource Paris Paris-6 "Networks & Infra" Allocation Télécom SystemX Our Contribution Dauphine University Research Team SudParis Conclusion and Future Work University Teaching : Combinatorial Optimization, Mathematical Programming,... VNF-FG Placement Game theory, Cloud, Openstack, NFV, SDN, C-RAN, 5G Main Contributions Conclusion Combinatorial Optimization, Graph theory, Linear algebra Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 2 / 69

Outline Contributions to Cloud and NFV Areas Makhlouf Research Context Hadji 1 Research Context 2 Virtual Machines Placement and Repacking Optimization VMs Placement & 3 Critical Resource Allocation in Cloud Federations Repacking Our contribution Conclusion and Further Research Virtualized Network Functions Chaining and Placement 4 Critical Problems Resource Allocation Our Contribution 5 Open Research Challenges Conclusion and Future Work VNF-FG Placement Main Contributions Conclusion Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 3 / 69

Resource Allocation in Cloud Data Centers Contributions Context : to Cloud and NFV Areas How/Where to deploy/reallocate optimally a set of VMs Makhlouf in a cloud data center ? Hadji Research Context VMs Placement & Repacking Placement Our contribution Conclusion and Cloud Further Research Manager Critical Resource Allocation Our Contribution Conclusion and Future Work VNF-FG Placement Main Contributions Conclusion Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 4 / 69

Critical Resource Allocation in Cloud Federations Contributions Context : to Cloud and NFV Areas Where to place interconnected VMs in a federation Makhlouf taking into account the importance of certain services ? Hadji Research Context VMs Placement & Placement Repacking Our contribution Conclusion and Further Research Prov i Critical Prov j Resource Allocation Our Contribution Conclusion and Future Work VNF-FG Cloud Federation Placement Main Contributions Conclusion Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 5 / 69

Network Resource Allocation Contributions Context : to Cloud and NFV Areas What is the optimal placement of VNFs chains in the Makhlouf NFV context ? Hadji Research Context VNF-FG (SFC) Flow 1 f 2 VMs Placement & Repacking f 1 Our contribution Flow 2 Conclusion and Further Research f 3 Critical Optimal Mapping Resource Allocation Our Contribution Conclusion and Future Work VNF-FG Placement Main Contributions Physical substrate Conclusion Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 6 / 69

Outline Contributions to Cloud and NFV Areas Research Context 1 Makhlouf Hadji Research Virtual Machines Placement and Repacking Optimization 2 Context Our contribution VMs Conclusion and Further Research Placement & Repacking Our contribution Conclusion and 3 Critical Resource Allocation in Cloud Federations Further Research Critical Resource Virtualized Network Functions Chaining and Placement 4 Allocation Problems Our Contribution Conclusion and Future Work Open Research Challenges VNF-FG 5 Placement Main Contributions Conclusion Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 7 / 69

Motivation Contributions to Cloud and NFV Areas Makhlouf Hadji To achieve increasing revenues, cloud providers require Research Context very efficient resource utilization and the strict respect of VMs quality of service, Placement & Repacking Without virtual resources (VMs) consolidation, initial Our contribution Conclusion and smart placement alone can not efficiently reduce costs Further Research Critical since all expenses matter : hosting, energy consumption, Resource Allocation maintenance, configuration and management costs Our Contribution Conclusion and Future Work VNF-FG Placement Main Contributions Conclusion Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 8 / 69

Motivation Contributions to Cloud and NFV Areas Makhlouf Definitions Hadji SLA Violation : is represented by the percentage Research Context of over used servers VMs Placement & Repacking Focus of our Work Our contribution Conclusion and Further Research Online placement and re-packing algorithms to Critical Resource reduce overall cost and improve resource Allocation utilization and, at the same time, minimize SLA Our Contribution Conclusion and violations Future Work VNF-FG Placement Main Contributions Conclusion Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 9 / 69

Problem Statement Contributions to Cloud and NFV Areas Description Makhlouf Hadji The Figure depicts the Research placement problem of N Context VMs on K available ser- VMs vers. Cloud services are Placement & Repacking characterized by elastic Our contribution requirements from users, Conclusion and Further Research that induce variable Critical workloads on the system Resource Allocation that requires conso- Our Contribution lidation to use more Conclusion and Future Work efficiently resources. VNF-FG Placement Main Contributions Conclusion Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 10 / 69

Problem Complexity Contributions to Cloud and NFV Areas Makhlouf Hadji Complexity Research Context If we consider each VM as an item and each available VMs server as a bin, then our VMs placement problem is Placement & Repacking very similar to the well known NP-Hard problem of the Our contribution Bin-Packing. Thus, our problem is also NP-Hard. Conclusion and Further Research We investigate scalable and cost-efficient algorithms to Critical Resource cope with large problem instances in negligible times. Allocation Our Contribution Conclusion and Future Work VNF-FG Placement Main Contributions Conclusion Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 11 / 69

Mathematical Modeling Contributions to Cloud and NFV Areas Bipartite Graph Construction Makhlouf Hadji If a VM i is currently Research VM i VM i Context hosted by a server k , VMs   then we consider W  W H R e H Placement & e ik ik ik Repacking hosting costs w ik = H ik . Our contribution Conclusion and Reconfiguration costs Server k Server k’ Server k Server k Further Research ( R ik ) are added if the Critical Resource VM i should be Allocation Our Contribution migrated from a server Conclusion and Future Work k ′ to a server k VNF-FG ( w ik = H ik + R ik ) Placement Main Contributions Conclusion Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 12 / 69

Mathematical Modeling Contributions to Cloud and Bipartite Graph Construction NFV Areas Makhlouf Hadji There is an edge H S VM 11 1 e = ( i , k ) between each 1 H  Research 12 R Context 12 VM i and each available VM S H 2 2 VMs 1  server k K Placement & R Repacking 1 K The weight of each edge Our contribution Conclusion and e = ( i , k ) is provided Further Research Critical S according to the current VM K Resource N Allocation placement of VM i Our Contribution V S Conclusion and Future Work VNF-FG Placement Main Contributions Conclusion Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 13 / 69

b-Matching Approach Contributions to Cloud and NFV Areas Proposition Makhlouf Hadji Let G = ( V ∪ S , E ) be a complete weighted bipartite Research Context graph, where V represents a set of Virtual Machines, VMs and S a set of physical servers hosting the VMs. Finding Placement & Repacking an optimal VMs repacking solution is equivalent to the Our contribution polynomial time solution of the uncapacitated minimum Conclusion and Further Research weight b-Matching problem, such as : Critical Resource b ( v ) = 1 , ∀ v ∈ V Allocation Our Contribution � � CPU v Conclusion and b ( v ) = min | V | ; ⌈ e ∈ δ ( v ) CPU I ( e ) ⌉ , ∀ v ∈ S Future Work � VNF-FG Placement Main Contributions Conclusion Research Challenges Makhlouf Hadji Contributions to Cloud and NFV Areas 3 March, 2017 14 / 69