Dynamic Resource Allocation in Clouds: Smart Placement with Live Migration

(1)

Dynamic Resource Allocation in

Clouds: Smart Placement with Live

Migration

Makhlouf Hadji

Ingénieur de Recherche [email protected]

(2)

FONDATION DE COOPERATION SCIENTIFIQUE

(3)

NOS VALEURS

❶

E

xcellence

-

Talents

-

Résultats

-

Echanges

❷

S

ouplesse

-

Fonctionnement

-

Partenariat

-

Projets

❸

R

igueur

-

Propriété Intellectuelle

-

Confidentialité

-

Exécution

(4)

PROGRAMMES DE R&D

4

(5)

PROJETS R&D

Projets opérationnels depuis le lancement

Projets Opérationnels M€ Financement Industriel

9 15,5

ETP/an sur 3 ans Partenaires Industriels Partenaires Académiques

110

42

12 Thèses

29

(6)

(7)

Smart Placement in Clouds

Define the best strategy to place VMs workloads leading to optimally reduce infrastructure costs.

ESX 1 ESX 2… ESX N

Infrastructure servers

???

Cloud End-Users VMs demand management VMs placement strategy Problem: Based on allocating and hosting N VMs on a physical

infrastructure of X Serveurs, what is the best manner to optimally place workloads to minimize different infrastructure costs ?

Benefits

 Resources optimization,

 Minimization of infrastructure costs,

 Energy consumption optimization.

VM Placement problem

ESX 1 ESX 2… ESX N

Non optimal placement

ESX 1 ESX 2… ESX N

Optimal placement

Challenges of the problem:

(8)

Smart Placement in Clouds



French Providers Point of View

(9)

Smart Placement in Clouds

ESX 1 ESX 2… ESX N

Infrastructure serveurs Utilisateurs des services Cloud Gestion de la demande de VM Forcasting & Scheduling s m al l large

Due to fluctuations in

users’ demands, we

use Auto-Regressive

(AR(k)) process, to

handle with future

demands:

Problem Complexity :

NP-Hard Problem: One can construct easily a plynomial reduction from the NP-Hard notary problem of the Bin-Packing. t i t k i i t

d





_









1

(10)

Smart Placement in Clouds

Mathematical formulation:

10

Formulation as ILP:

The corresponding mathematical model is an Integer Linear Programming: difficulties to characterize the convex hull of the considered problem and the optimal solution.































    

else.

0 i

server

in

hosted

is

VM

if

1 ,

,

1 ,

,

:

To

Subject

min

1 1 1 1 1 j ij ij j N i ij ij ij ij N i I j N i I j ij j ij ij

y

j

i

N

x

I

j

d

x

N

i

I

j

y

C

x

P

y

Z



(11)

Smart Placement in Clouds

Minimum Cost Maximum Flow Algorithm

S

T

(2; 0,23)

Instance i

(12)

Smart Placement in Clouds

Minimum Cost Maximum Flow Algorithm

12

S

T

(2; 0,23) Small Instance (2; 0,23) Medium Instance

(13)

Smart Placement in Clouds

Simulation Tests: Case of (0;1) Random Costs

We consider (0; 1) Random hosting costs between each couple of vertices (a, b), where a is a fictif node, and b is a physical machine (server).

(14)

Smart Placement in Clouds

Simulations Tests: Case of Inverse Hosting Costs:

14

Inverse Hosting Costs Scenario

We consider inversed hosting costs function between each couple of vertices (a, b), where a is a fictif node, and b is a physical machine:

Where represents the available capacity on the considered arc. est une fonction non nulle.









_ab _ab ab ab

C

g

C

f

g

if

0 ,

otherwise

)

(

1

ab

C

f

(15)

(16)

Live Migration of VMs



Migration process:

16

ESX

Xen

KVM

Hyper-V

(17)

Polytops, faces and facets

facets

face



c

j

x

j

M in



a

_ij

x

_j



b

_i

,

i



1 ,...,

m

Subject To constraints:

 

i

n

x

i



0 ,

1 ,



1 ,...,

*

x

Live Migration of VMs

Polyhedral Approachs 1 1





x



2 2





x



(18)

Live Migration of VMs



Decision Variables:

if

A VM k is migrated from i to j (0 else).



Prevent backword migration of a VM:



Server’s destination uniqueness of a VM migration:



Servers’ power consumption limitation constraints:



Etc…

Polyhedral Approachs

18

Some Valid Inequalities of our Problem:

1 

ijk

Z

1

'





_jlk ijk

Z

1

' , 1





  m i j j ijk

Z



j j current



j



ijk m i k qi k

y

p

z

p







 

1

, max , ' 1 1

(19)

Live Migration of VMs

























































 



  

            

idle

is

server

if

1 otherwise

0 j,

to

i

from

migrated

is

VM

if

1 '

1

1 '

,

'

,

1 ,

,

1 '

,

1 ,

'

,

1 ,

1 :

To

Subject

'

max

0 ' 1 ,max ' 1 , ' 1 1 ' 1 1 , max , ' , 1 ' ' 1 ' 1 ' 1 1 ,

i

y

z

T

t

z

P

m

y

q

z

y

P

z

p

z

k

j

l

m

l

q

k

q

k

m

i

j

m

i

z

p

y

P

M

k ijk k ijk m i j m j cu rren t j i m j i i q k ijk m i q k j cu rren t j j ijk k m i j j ijk j i jlk ijk m i m i m j q k ijk k i id l i i i i