Optimal Provisioning for Elastic Service Oriented Virtual Network

Optimal Provisioning for Elastic Service Oriented Virtual Network Request in Cloud Computing 101062558 劉冠逸 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Outline Introduction n Problem description n Genetic Algorithm-based Heuristic Algorithm (GAH) n Simulations n 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Introduction n Cloud computing paradigm enables users to access services and applications hosted in data centers based on their requirements. n The service or application request submitted to a data center can be abstracted as a virtual network (VN) request, which consists of a set of VN nodes and VN edges. 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Virtual Network 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Introduction n How to efficiently provision VN requests in multi datacenters is of utmost importance n For the elastic resource requirement services, providers need to make sure the Qo. S or SLAs are satisfied. 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Problem description (I) 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Problem description (II) n 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Problem description (III) n The revenue R(GV) generated by provisioning a VN request can be calculated as follows: 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Greedy VN Provisioning Algorithm(GVNP) n sss 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Greedy VN Provisioning Algorithm(GVNP) n sss 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Genetic Algorithm-based Heuristic Algorithm (GAH) Chromosome Coding n Chromosome Operations n Genetic Algorithm-based Heuristic Algorithm (GAH) n 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Chromosome Coding The number of columns in the array equals to the number of server nodes in substrate network n The total number of element “ 1” in the array equals to the number of VN nodes in a VN request n 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Chromosome Operations n n n Cloning Crossover Mutation Feasibility checking Selection 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Chromosome Operations n Cloning l l The cloning operation involves generating theinitial population The GA procedure begins iterations from this population 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Chromosome Operations n Crossover 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Chromosome Operations n Mutation l l The mutation operation is used to prevent solutions from being trapped at a local optimum Mutation is done in the children population, by changing the values of some genes with a small probability p (from 0. 001 to 0. 1) m 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Chromosome Operations n Feasibility checking l Some of the newly generated children may not be feasible solutions for the original problem. 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Chromosome Operations n Selection l The chromosome selection is to select parent chromosomes from the particular generation of population, and assign reproductive opportunities to these selected chromosomes 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Genetic Algorithm-based Heuristic Algorithm (GAH) 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Simulations n Use the ITALYNET (Figure 4) with 20 nodes and 36 links as substrate network in our simulation 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Simulations 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Simulations 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

Conclusion n In this work we address the problem of optimal provisioning for elastic service oriented VN request in cloud-based datacenters. n We model this problem as a mathematical optimization problem by using mixed integer programming and propose a genetic algorithm based heuristic algorithm for solving this NP-hard problem efficiently. n The experimental results demonstrate that the solution obtained by our approach is near to the optimal solution 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory

The End 國立清華大學高速通訊與計算實驗室 NTHU High-Speed Communication & Computing Laboratory