8. Disaster Mitigation Needs
We have faced devastating disasters in the last years
Tsunamis
Japan 2011
Indonesia and Thailand 2004
Earthquakes
Haiti 2010
China 2010
Chile 2010
Hurricanes
Katrina August 2005
Wilma October 2005
9. Disaster Mitigation Needs
Better prediction capacity is required to generate better
evacuation plans
Finer grained predictions
Specific geographical area
Predictions should be obtained in time
Federated support vs Centralised support
10. Disaster Mitigation Needs
Prediction models usually require a large amount of
computing power
Grid and Cloud Computing can potentially bring more
power than Clusters
12. The WRF Web Portal
The Weather Research and Forecasting (WRF)
Gaining acceptance as one of the best models
Can also be used to predict evolution of forest fires and predict
air pollution
However, demands a huge amount of CPU power
13. The WRF Web Portal
WRF Web Portal
On demand runs of WRF
Available for meteorologist of all Mexican States
Support for:
More accurate weather prediction on a local area
Research on meteorology
Private cloud computing environment
Challenge
Increase performance while retain benefits of cloud computing
environment
15. Cloud Computing and HPC
Examples of HPC applications include:
Weather Forecasting
Quantum Chemistry Calculations
Accelerator Modelling
Astrophysical Simulations
This kind of Apps have huge demands of computing
resources
Traditionally supported by cluster- or Grid-based systems
16. Cloud Computing and HPC
Has experienced a tremendous growth recently
Offers a number of advantages over other distributed paradigms.
Access to resources is
Flexible
cost-effective. No need to pay large amount of money for
Infrastructure
Salaries for maintenance functions
The possibility of using Cloud computing for HPC Apps is
attractive
17. Cloud Computing and HPC
Current cloud computing platforms are not suitable for
running some of these applications
The performances offered is very poor due to the extensive
use of virtualisation
The performance is specially degraded when there is a huge
amount of IO operations
No guarantees for QoS (i.e. execution time) can be offered
18. Performance Prediction
Allows for Capacity planning
Helps to establish and maintain a Service Level Agreement
(SLA)
Most approaches either focus on
a specific application
a specific platform configuration
19. Performance Prediction
Need for supporting the following characteristics
Application agnostic
Platform configuration independency
Multicore systems
Virtualised environments
20. Conclusions
We presented the LA Grid initiative
We have identified some needs in natural disaster
mitigation
The WRF Web Portal case example was presented
Some challenges were outlined
21. Thank you!
Hector Duran-Limon
University of Guadalajara, Mexico
hduran@cucea.udg.mx
http://maestro.cucea.udg.mx/~hduran/