grid based environment application development -...

SEE-GRID-SCI

SEE-GRID-SCI Dissemination - Timisoara 26 September 2009

Grid Based Environment Application Development - GreenView

Danut Mihon, Victor Bacu, Teodor Stefanut, Dorian GorganTechnical University of Cluj-Napoca

{vasile.mihon, victor.bacu, teodor.stefanut, dorian.gorgan}@cs.utcluj.ro

SEE-GRID-SCI Dissemination 26 September 2009


GreenView objectives

It is an environmental application used for predicting and monitoring the temperature over a specific geographical area

In present it works only for CEE regionsExtend it’s functionality to other geographical areas

Calibrate satellite measurements Based on BIOME-BGC (BioGeochemical Cycles) model

Integrate other environmental applications as subcomponents of GreenViewEnviMon (RENAM)Soil Pollution by Heavy Metals (CENS NAS RA - Center for Ecological-Noosphere Studies of National Academy of Sciences of the Republic of Armenia)

The application provide a quick and easy access to it’s functionality based on a web interface

http://gisheo01.mediogrid.utcluj.ro:8095/interpolation_v2.1/http://wiki.egee-see.org/index.php/GreenView


GreenView related architecture


Why we choose these technologies?

Server side algorithms are implemented using Java SDK 1.5

The user interface is created using Adobe Flex 3.3 technology and BlazeDS, that allow free access to remote technologies developed by Adobe.

Adobe Flex advantagesFree technologyAllows RIA developmentUses MXML components for GUI and action script for interactivityPredefined componentsAdd dynamic behaviorPlatform independentBuild the code into a SWF that runs in Flash Player


GreenView – Application Development Methodology

1. Algorithm identification and analysis2. Data model definition3. Identify atomic parts of the algorithms

– Parallel or serial processing– Implementation as services, procedures, separate applications

4. Algorithm implementation5. Description of processing using gProcess

– PDG Definition– iPDG specification– Execution

6. Create the application interface– Link the application interface to the GRID level


GreenView - Algorithm identification and analysis

Interpolation: eg. neighbor interpolation, bicubic interpolation, bilinear interpolation etc.Solution: Coarse to fine interpolation

The missing points will be computed using a distance weighted nonlinear interpolation that computes the value of a Vm point from four surrounding known values: V1, V2, V3 and V4.

Vi - one of the four surrounding pixels, Vi( ); di - the great circle distance between two points; dmax - the great circle distance between the furthest two pixels of the surrounding four pixels; R - average radius for a spherical approximation of the Earth (≈6371.01Km)


GreenView - Algorithm implementation

Execute operation on Grid infrastructure uses gProcesss platformThe entire GreenView application is divided into atomic components (jobs)Each job is executed on different nodes of the GridCreate a PDG (Process Description Graph) and an iPDG (instantieted PDG) that specify the resources to be used on execution timeThe execution ends when the job reaches DONE statusThe result is copied to the server, in a database

Example of a PDG for coarse to fine component


Coarse to fine – performance evaluation


GPP computation - performance evaluation


Conclusions on the experiments

On the Grid infrastructure the time execution of a job means

Time to search free nodes of the Grid

Time to send the required information

Time required to execute the job

Time required to retrieve the results from the Grid

The gap that appears in these charts is determined by

Time for searching free nodes on each site

The computation power of a node

The network traffic and bandwidth

Only the applications that require large processing data volume and the applications that are using the parallel computing power of the Grid infrastructure should be implemented as Grid applications


Coarse to fine interpolation – input/output example

Left image is the coarse one, with lower resolution

The result represents the same geographical area as the first one, but at a higher resolution.


Coarse to fine – step 1


Fine to coarse – step 1


GPP computing– step 1


About GreenView – step 1


Conclusions

GreenView is a useful tool for predicting and monitoring the temperatures over geographical areas

Even though it works, for now, only on CEE regions, the application can be extended globally

Could integrate another environment applications, such as forecast prediction

GreenView provides quick and easy access to it’s functionality through a web interface


Acknowledgments

This research is supported by SEE-GRID-SCI (SEE-GRID eInfrastructure for regional eScience) project, funded by the European Commission through the contract nr RI-211338.

Climate change data have been retrieved from the PRUDENCE data archive, funded by the EU through contract EVK2-CT2001-00132.

MODIS data have been produced and distributed by NASA through the EOS Data Gateway system.

Biome-BGC version 4.1.1 was provided by the Numerical Terradynamic Simulation Group (NTSG) at the University of Montana. NTSG assumes no responsibility for the proper use of Biome-BGC by others.


Thanks for attending!

grid based environment application development -...

Documents