Dynamic Resource Allocation in Clouds: Smart Placement with Live Migration



Similar documents
Capacity Planning for Virtualized Servers

Stochastic Models of Load Balancing and Scheduling in Cloud Computing Clusters

Stochastic Models of Load Balancing and Scheduling in Cloud Computing Clusters

Stochastic Models of Load Balancing and Scheduling in Cloud Computing Clusters

Virtual machine resource allocation algorithm in cloud environment

An Electricity Trade Model for Microgrid Communities in Smart Grid

Online Algorithms for Uploading Deferrable Big Data to The Cloud

BANDWIDTH ALLOCATION AND PRICING PROBLEM FOR A DUOPOLY MARKET

Inventory Control in a Multi-Supplier System

A Multi Due Date Batch Scheduling Model. on Dynamic Flow Shop to Minimize. Total Production Cost

The Greedy Method. Introduction. 0/1 Knapsack Problem

How To Write A Powerpoint Powerpoint Commandbook For A Data Center

Hosting Virtual Machines on Distributed Datacenters

Research Article Load Balancing for Future Internet: An Approach Based on Game Theory

Least Squares Fitting of Data

Formulating & Solving Integer Problems Chapter

Self-Adaptive Capacity Management for Multi-Tier Virtualized Environments

Multi-timescale Distributed Capacity Allocation and Load Redirect Algorithms for Cloud System

Basic Queueing Theory M/M/* Queues. Introduction

Naglaa Raga Said Assistant Professor of Operations. Egypt.

Network Aware Load-Balancing via Parallel VM Migration for Data Centers

Loop Parallelization

The Packing Server for Real-Time Scheduling of MapReduce Workflows

Self-Adaptive SLA-Driven Capacity Management for Internet Services

An Analytical Model of Web Server Load Distribution by Applying a Minimum Entropy Strategy

INSTITUT FÜR INFORMATIK

II. THE QUALITY AND REGULATION OF THE DISTRIBUTION COMPANIES I. INTRODUCTION

Fragility Based Rehabilitation Decision Analysis

Support Vector Machines

GRAVITY DATA VALIDATION AND OUTLIER DETECTION USING L 1 -NORM

CONSTRUCTION OF A COLLABORATIVE VALUE CHAIN IN CLOUD COMPUTING ENVIRONMENT

Tourism Demand Forecasting by Improved SVR Model

A Statistical Model for Detecting Abnormality in Static-Priority Scheduling Networks with Differentiated Services

Jimmy Hébergement Cloud - TechDay

Feature selection for intrusion detection. Slobodan Petrović NISlab, Gjøvik University College

Resource Management For Workflows In Cloud Computing Environment Based On Group Technology Approach

LITERATURE REVIEW: VARIOUS PRIORITY BASED TASK SCHEDULING ALGORITHMS IN CLOUD COMPUTING

Virtual Network Embedding with Coordinated Node and Link Mapping

CLoud computing technologies have enabled rapid

2008/8. An integrated model for warehouse and inventory planning. Géraldine Strack and Yves Pochet

Maximizing profit using recommender systems

Project Networks With Mixed-Time Constraints

Introduction CONTENT. - Whitepaper -

Scan Detection in High-Speed Networks Based on Optimal Dynamic Bit Sharing

Minkowski Sum of Polytopes Defined by Their Vertices

How To Calculate An Approxmaton Factor Of 1 1/E

POLYSA: A Polynomial Algorithm for Non-binary Constraint Satisfaction Problems with and

Elastic Systems for Static Balancing of Robot Arms

Netherlands Published online: 27 Jun 2013.

A method for a robust optimization of joint product and supply chain design

An MILP model for planning of batch plants operating in a campaign-mode

INVENTORY CONTROL FOR HIGH TECHNOLOGY CAPITAL EQUIPMENT FIRMS. Hari Shreeram Abhyankar. B.S. Mathematics B.S. Economics Purdue University.

Economic Models for Cloud Service Markets

Fair Virtual Bandwidth Allocation Model in Virtual Data Centers

J. Parallel Distrib. Comput.

Allocating Collaborative Profit in Less-than-Truckload Carrier Alliance

Efficient On-Demand Data Service Delivery to High-Speed Trains in Cellular/Infostation Integrated Networks

CALL ADMISSION CONTROL IN WIRELESS MULTIMEDIA NETWORKS

A Novel Dynamic Role-Based Access Control Scheme in User Hierarchy

Fisher Markets and Convex Programs

OPTIMAL INVESTMENT POLICIES FOR THE HORSE RACE MODEL. Thomas S. Ferguson and C. Zachary Gilstein UCLA and Bell Communications May 1985, revised 2004

8 Algorithm for Binary Searching in Trees

8.5 UNITARY AND HERMITIAN MATRICES. The conjugate transpose of a complex matrix A, denoted by A*, is given by

Extending Probabilistic Dynamic Epistemic Logic

GENERALIZED PROCRUSTES ANALYSIS AND ITS APPLICATIONS IN PHOTOGRAMMETRY

How To Trade Water Quality

n + d + q = 24 and.05n +.1d +.25q = 2 { n + d + q = 24 (3) n + 2d + 5q = 40 (2)

Section 2 Introduction to Statistical Mechanics

Price Competition in an Oligopoly Market with Multiple IaaS Cloud Providers

Availability-Based Path Selection and Network Vulnerability Assessment

Resource Allocation Model to Find Optimal Allocation of Workforce, Material, and Tools in an Aircraft Line Maintenance

Transcription:

Dynac Resource Allocaton n Clouds: Sart Placeent wth Lve Mgraton Mahlouf Had Ingéneur de Recherche ahlouf.had@rt-systex.fr Avec : Daal Zeghlache (TSP) daal.zeghlache@teleco-sudpars.eu

FONDATION DE COOPERATION SCIENTIFIQUE 2

NOS VALEURS ❶ Excellence - Talents - Résultats - Echanges ❷ Souplesse - Fonctonneent - Partenarat - Proets ❸ Rgueur - Proprété Intellectuelle - Confdentalté - Exécuton 3

PROGRAMMES DE R&D Ingénere nuérque des Systèes et Coposants 4

PROJETS R&D Proets opératonnels depus le lanceent 9 15,5 Proets Opératonnels M Fnanceent Industrel 110 42 12 ETP/an sur 3 ans Partenares Industrels Partenares Acadéques 29 Thèses 5

I- Sart Placeent n Clouds

Sart Placeent n Clouds VM Placeent proble Proble: Based on allocatng and hostng N VMs on a physcal nfrastructure of X Serveurs, what s the best anner to optally place worloads to nze dfferent nfrastructure costs? VMs deand anageent Cloud End-Users ESX 1 ESX 2 ESX N Non optal placeent VMs placeent strategy??? Optal placeent ESX 1 ESX 2 ESX N ESX 1 ESX 2 ESX N Benefts Resources optzaton, Mnzaton of nfrastructure costs, Energy consupton optzaton. Challenges of the proble: Exponental nuber of cases to enuerate. Infrastructure servers Defne the best strategy to place VMs worloads leadng to optally reduce nfrastructure costs. 7

Sart Placeent n Clouds French Provders Pont of Vew 8

Sart Placeent n Clouds Due to fluctuatons n users deands, we use Auto-Regressve (AR()) process, to handle wth future deands: d t 1 d t t Geston de la deande de VM sall Forcastng & Schedulng Utlsateurs des servces Cloud large ESX 1 ESX 2 ESX N Proble Coplexty : NP-Hard Proble: One can construct easly a plynoal reducton fro the NP-Hard notary proble of the Bn- Pacng. Infrastructure serveurs 9

Sart Placeent n Clouds Matheatcal forulaton: Forulaton as ILP: The correspondng atheatcal odel s an Integer Lnear Prograng: dffcultes to characterze the convex hull of the consdered proble and the optal soluton. n Z 1 d Subect To: x x N y x C y, I N,, 1f VM 0 else., I, 1 1 1 1 N I y N 1, N s hosted n server I P x 10

Sart Placeent n Clouds Mnu Cost Maxu Flow Algorth Instance (2; 0,23) S T Legend: (capacty; cost) 11

Sart Placeent n Clouds Sall Instance Mnu Cost Maxu Flow Algorth (2; 0,23) Medu Instance S T (2; 0,23) 12

Sart Placeent n Clouds Sulaton Tests: Case of (0;1) Rando Costs Rando Hostng Costs Scenaro We consder (0; 1) Rando hostng costs between each couple of vertces (a, b), where a s a fctf node, and b s a physcal achne (server). 13

Inverse Hostng Costs Scenaro Sart Placeent n Clouds Sulatons Tests: Case of Inverse Hostng Costs: We consder nversed hostng costs functon between each couple of vertces (a, b), where a s a fctf node, and b s a physcal achne: g 1 f C ab ab ab f ( Cab) Where C represents the avalable capacty on the consdered arc. est une foncton non nulle. ab f 0, otherwse g 14

II- Lve Mgraton of VMs

Lve Mgraton of VMs Mgraton process: Xen ESX KVM Hyper-V 16

Lve Mgraton of VMs Mn Polytops, faces and facets c x Subect To constrants: a xb, 1,..., 0,1, 1 n x,..., x 1 1 Polyhedral Approachs face x* facets x 2 2 20

Lve Mgraton of VMs Soe Vald Inequaltes of our Proble: Polyhedral Approachs Decson Varables: Z 1 Prevent bacword graton of a VM: f A VM s grated fro to (0 else). Z Z l 1 Server s destnaton unqueness of a VM graton: Z 1, 1 Servers power consupton ltaton constrants: Etc q 1 1 p z p p 1 y,ax, current 18

Lve Mgraton of VMs Polyhedral Approachs 19 0 otherwse dle s server 1f 0 otherwse to, fro grated s VM 1f 1 1,, 1,, 1,, 1,,,, 1, 1, Subect To: ax 0 1 ax, 1, 1 1 1 1,,ax 1, 1 1 1 1, y z T t z P P y y q z y P P z p z l l q q z z z p y P M current q q current l q dl

Lve Mgraton of VMs Polyhedral Approachs Nuber of used servers when tang nto account Mgratons 20

Convergence Te (n seconds) of Mgraton Algorth: 21

Lve Mgraton of VMs Polyhedral Approachs Percentage of Ganed Energy when Mgraton s Used 5 10 15 20 25 30 10 35,55 36,59 00 00 00 00 20 27,29 34,00 35,23 38,50 00 00 30 17,48 27,39 35,21 40,32 41,89 36,65 40 16,77 18,85 22,02 32,31 39,90 40,50 50 10,86 16,17 19,85 22,30 39,20 36,52 60 08,63 14,29 18,01 22,13 25,15 30,68 70 08,10 14,00 14,86 15,90 22,91 23,20 80 07,01 10,20 10,91 15,34 17,02 21,60 90 06,80 09,32 10,31 14,70 16,97 19,20 100 05,90 07,50 08,40 12,90 16,00 14,97 22

Than you