12
CloudERT Pilot

Demo cloud ert_withoutvideos

Embed Size (px)

Citation preview

Page 1: Demo cloud ert_withoutvideos

CloudERT Pilot

Page 2: Demo cloud ert_withoutvideos

Objectives

• Cancer is Europe’s second largest killer• is a suite of remote tools to help medical physicists in the

definition of treatment plans and their verification using MC• This experiment focuses on the feasibility of using Cloud for such

services, using treatment verification as an example.– CloudERT must produce results ASAP by exploiting the Cloud resources.– Transferring the computing to Cloud

• User Community– Medical physicists– 65 users from 47 hospitals

Page 3: Demo cloud ert_withoutvideos

Software Design

• Steps 2 and 4 in parallel• Programmed used COMPSs and GW

COMPSs programming model

Page 4: Demo cloud ert_withoutvideos

Software Design

• GW Programming mode

App-specificApp-specificClients to Standard VENUS-C

Clients to Standard VENUS-C

ExternalExternal

eIM

RTeI

MRT

Dat

a tr

ansf

er

Clie

ntD

ata

tran

sfer

Cl

ient

Client

Job

Subm

issi

on

Clie

ntJo

b Su

bmis

sion

Cl

ient Cloud

Storage

Data Access Service

Data Access Service

Application Repository

Application Repository

Alignment worker

Alignment workerAlignment

workerAlignment

workerAlignment worker

Alignment worker

BEAMnrcBEAMnrc

CloudClo

udLocal Drive

wra

pper

wra

pper

wra

pper

wra

ppe

r

Execution Service

Execution Service

Tver 1Tver 1wra

pper

Tver 3Tver 3Local Drive

wra

pper

Tver 2Tver 2

wra

ppe

r

Alignment worker

Alignment workerAlignment

workerAlignment

workerAlignment worker

Alignment worker

DOSXYZnrcDOSXYZnrc

CloudDri

CloudDri

CloudDri

Local Drive

wra

ppe

rw

rapp

erw

rapp

erw

rapp

erVENUS-C Standard

component

VENUS-C Standard

component

Page 5: Demo cloud ert_withoutvideos

Benefits of using Venus-C

• Computing power -> Cloud• Reducing the entry cost. Business opportunity.• Scales well with the problem size, reducing the time

to solve a single case.• The capability of choosing the number of resources

used and change them dynamically.

Page 6: Demo cloud ert_withoutvideos

Lessons Learnt

• The GW & COMPSs frameworks fit well the execution model of eIMRT.

• A better logging and error management is needed.• Work in easing the deployment of the service side is

still needed, maybe by providing preconfigured images.

Page 7: Demo cloud ert_withoutvideos

Example Case

• A radiotherapist wants to verify a radiation therapy treatment. • He select the desired input files for the treatment• Finally he obtains the calculated dose• Synthetic case based on Carpet or Quasimodo

– single segment treatment– 300.000 BEAMnrc histories – 3.000.000 DOSXYZnrc histories. – 60.000 histories per task

• Input FilesFile Description SizeCesga.tar.gz Java application plus Linux scripts 17.8kBVenus-c-filesystem.gz Tar file with executables, libraries and other Monte Carlo data. 18.2 MBMapData.raw Patient’s CT in internal format 60.3MBRtplan.xml RTplan in XML format 20.6kBTomograph.ramp Data to convert CT from Hounsfield units to densities 337 BUseracc.config File with user’s specific parameters for the used LINAC 55B

Page 8: Demo cloud ert_withoutvideos

Restults

Phase COMPSs MS AzureTver1 15 15BEAMnrc (average) 795 445Tver2 38 21DOSXYZnrc (average) 533 316Tver3 24 17

Execution timesSample Calculated dose

Real case execution (100 tasks)

Page 9: Demo cloud ert_withoutvideos

Conclusions

• VENUS-C CloudERT is a client application consuming an eIMRT processing service deployed in the cloud.

• It uses a coarse-grain data-flow programming model successfully exposed by both GW and COMPSs. CDMI was tested

• The feasibility of using this framework in Monte Carlo radiotherapy simulations has been proved with a synthetic treatment verification

• Proof-of-concept of using on-demand remote computing capacity to bring innovative services to hospitals with radiotherapy facilities through Internet

Page 10: Demo cloud ert_withoutvideos

Future Plans

• Integrate the cloud execution within the eIMRT platform, both GW based and COMPSs

• Use the Accounting module in order to analyse the business model

• Execute the treatment optimization in the Cloud

Page 11: Demo cloud ert_withoutvideos

Additional information

• References– J. Pena, D. M. González-Castaño, F. Gómez, A. Gago-Arias, F. J.

González-Castaño, D. Rodríguez-Silva, A. Gómez, C. Mouriño, M. Pombar, and M. Sánchez. eIMRT: a web platform for the verification and optimisation of radiation treatment plans. In Journal of Applied Clinical Medical Physics , volume 10, 2009.

– A. Gómez, J.C. Mouriño, L.M. Carril, Z. Martín, D. Lezzi, R. Rafanell, and R.M Badía. Execution of Monte Carlo Treatment Verification on Cloud using COMPSs Platform. Proceedings of the 3rd European Workshop on Monte Carlo Treatment Planning (MCTP’12). pp. 186-189, 2012.

• URL for the eIMRT project– http://eimrt.cesga.es

Page 12: Demo cloud ert_withoutvideos

Contact

Carlos MouriñoCentro de Supercomputación de GaliciaAv. De Vigo s/n (Campus Vida),Santiago de Compostela 15705 A Coruña, SpainTel: +34-981569810Fax. +34-981594616E-mail: [email protected]