1

Performability Models for Designing Disaster Tolerant Cloud Computing Systems

2

Agenda

Motivation

Approach

Architecture

Perfomability

Sensitivity Analysis

Survivability Evaluation

GeoClouds Tool

3

MOTIVATION

Why adopt cloud computing?

4

Why adopt cloud computing?

Cloud computinginfrastructure is available on demand.

Adopted as a service.

Minimizes the costs of IT infrastructures

Service Level Agreement (SLA). Penalties may be applied if the defined availability level is not

satisfied

5

Motivation

Large cloud service providers adopts service level agreements (SLAs) to regulate the availability of the cloud service.Costs

Availability

Performance (response time)

Service provider needs to carry out availability analysis

6

Motivation

IaaS – computing resources in the form of virtual machines (VMs).

DisastersMultiple data located in different geographical locations

Availability improvement

VM migration time increases due to distance between data centers

7

APPROACH

8

APPROACH

Evaluate dependabilitycloud computing systems deployed into geographically distributed data

centers.

(RBD - Reliability Block Diagrams)

(SPN - Stochastic Petri Nets)

9

ARCHITECTURE

10

Architecture

11

(SPN/RBD) BUILDING BLOCKS

12

Hierarchical Modelling

13

COMBINED MODEL

14

Survivability Evaluation

15

RTO and RPO Evaluation

RTO Model

RPO Model

16

Results

17

Sensitivity Analysis

18

Results

19

Results

20

Geoclouds Modcs

21

Validation

22

Validation

23

Conclusion

• Performability

• Survivability

• Tool

• Validation

• Sensitivity Analysis

Next Steps :

Write Papers

24

Questions