Open Grid Computing Environments: Advanced
Gateway Support Activities
RT Project ReviewOctober 7th, 2010
Software Description
OGCE Gadget Container
Google Gadget/Open Social compatible software for building Web-based user interfaces.
XBaya A visual user interface for composing, launching and monitoring workflows
GFAC An application factory service for wrapping command-line tools as Web services
XRegistry A service and workflow registry
Registry Gadget, Experiment Builder
User interface for registering services and creating online experiments with registered workflows
Software Strategy• We develop and package downloadable software• Focus on gadget container and tools for running
science applications on grids and clouds.• Provide a tool set that can be used in whole or in
part.– If you just want GFac, then you can use it without
buying an entire framework.• Outsource security, information services, data
and metadata, etc to other providers.– MyProxy, TG IIS, Globus, Condor, XMC Cat, iRods, etc.
NSF SDCI Award: 2010-2013
• Indiana: Marlon Pierce (PI), Suresh Marru (Co-PI), Raminder Singh, Rob Quick, Gregor von Laszewski, Gerald Guo
• Purdue: Carol Song (Co-PI), David Braun, Lan Zhao
• NCSA/UIUC: Sudhakar Pamidighantam (Co-PI), Shaowen Wang, Yan Liu
• UTHSCSA: Emre Brookes (Co-PI), Borries Demeler
Project Highlights• Full Circle Development
– Directly fund both software developers and gateway consumers.
• Directly supported (non-IU) gateways: – UltraScan (UTHSCSA), GridChem (NCSA),
SimpleGrid/GISolve (UIUC), Purdue CCSM and Environmental Data Gateways
– Among the most used TG gateways.• Sustainability strategy: Apache Incubator for
workflow suite of tools – XBaya, GFac, and supporting services.
UTHSCSA JacintoTerascale storage
Web Server
US LIMS
MySQL DB
User
High Performance Computing Clusters
TeraGrid
TIGRE/Globus Network
GridControl
UltraScan High Level Overview
UltraScan Middleware
• Immediate Goals: Use GFAC as a replacement job submission service. – GRAM 2, 4, 5 independence– Significant effort into GRAM5
testing on Ranger.• Longer term goals
– Integrate with TG information services to provide better job scheduling.
• OGCE Resource Prediction Service– Support UNICORE job
management. Current Architecture
Gram5 Testing Steps• Developed Testing harness to run different
UltraScan cases.• Started with small number of jobs and increased
the concurrency later• Watched job behavior of the job on resource and
monitored the gram log– There were lot of issue which we found from
the logs and working with Globus team to fix them
• Recorded all the job run data to create a google gadget to create graph for different runs on different resources.
Patterns:
TG Resources and PatternsVersion Resource EndpointGT 5.0.2 QueenBee queenbee.loni-lsu.teragrid.org:2120/jobmanager-pbs
GT 5.0.2 Ranger login5.ranger.tacc.teragrid.org:2120/jobmanager-sge
GT 5.0.2 Lonestar gatekeeper.lonestar.tacc.teragrid.org:2120/jobmanager-lsf
Concurrent jobs Batch Size Total jobs Job Status Pass : Fail
1 10 10 10:0
3 10 30 30:0
5 10 50 50:0
10 10 100 20:0
20 10 200 40:0
50 10 500 100:0
100 10 1000 200:0
200 5 1000 Not tested (Need allocation)
500 2 1000 Not tested (Need allocation)
• There were problems with Sun Grid Engine job manager (sge.pm) not loading system environment and user environment settings.
• You can't load modules using Gram parameters. – On Ranger all the libraries are modules and user may want to load
some special module version for his job. – Example is if you have different version of your code using different
modules.
• On Ranger user default shell is C shell and Gram runs using bash shell. This is not a problem but users need to be informed where to add user environment.
• You can look at detailed results on followinghttp://teragridforum.org/mediawiki/index.php?title=GRAM_5_Testing
Learning from this exercise on Ranger
GridChem Science Gateway• A chemistry/material Science Gateway for running
computational chemistry codes, workflows, and parameter sweeps.
• Integrates molecular science applications and tools for community use.
• 400+ users heavily using TeraGrid. One of the consistent top5 TeraGrid Gateway users.
• Supports all popular Chemistry applications including Gaussian, GAMESS, NWChem, QMCPack, Amber and MolPro, CHARMM
• ParamChem is a follow-on project to develop workflows for chemical parameter studies and provide the infrastructure to execute them.
Cyberenvironments for ParameterizationComputational Reference Data Generation
GridChem , ParamChem Integration
• Immediate Goals: integrate XBaya workflow composer with GridChem user interface and middleware.– Target SC10
• Longer term goals: provide part of GridChem next generation middleware; support large scale ParamChem workflows.
SimpleGrid (UIUC)
SimpleGrid is a Gateway teaching tool that is also the core of GISolve
SimpleGrid Gadgets
Yan and Gerald’s work from our Oct 6th hacking session.
SimpleGrid, GISolve
• Short term goal: develop SimpleGrid Gadgets deployable into gadget container.– Must meet security requirements– Support PHP development– Support interactivity requirements
• Integrate YUI JavaScript libraries with Gadget JavaScript.
• Longer term goals: investigate workflow, job management tools. Apply to GISolve
Purdue CCSM and Data Portals
• Short terms goals: Develop CCSM and data management gadgets and necessary backing middleware.– Support GWT development– Interactivity and security requirements.– Significant requirements overlap with SimpleGrid
• Longer term goals: Build gateways out of gadgets hosted by multiple containers; examine workflow and other tools.
Conclusion• We focus initially on one component per
gateway.– SimpleGrid, CCSM, Data Portal: gadgets
• Other gadget based gateways at UC– GridChem: Xbaya– UltraScan: GFac
• Goal is to establish an Apache-style meritocracy for contributed code.
• Making distributed teams work: hacking retreats, SourceForge, Jira, Google apps
Acknowledgments
• The following student interns have contributed to the project over the last year– Ye Fan– Patanachai Tangchaisin– Suresh Kumar Deivasigamani– Roland Mai– Sashikiran Challa
Backup Slides
2121
OVP/RST/ MIG
OGCERe-engineer, Generalize,
Build, Test and Release
LEAD
OGCE Gateway Tool Adaption & Reuse
GridChem
TeraGridUser Portal
OGCE Team
GridChem
Ultrascan
BioVLab
ODI
Bio Drug Screen
EST Pipeline
Future Grid
GFac, XBaya, XRegistry, FTR
Eventing System
LEAD
Resource Discovery Service
GPIR, File Browser
Gadget Container, GTLab, Javascript Cog,
XRegistry Interface, Experiment Builder, Axis2 Gfac, Axis2 Eventing System,
Resource Prediction Service, Swarm
Experiment Builder, XRegistry Interface
Xbaya, GC Middleware
GFac, Eventing System
XBaya, GFac
Workflow Suite, Gadget Container
Swarm->GFac
Swarm->GFac
GFac, Xbaya, …
Putting It All Together
Software Strategy
• Focus on gadget container and tools for running science applications on grids and clouds.
• Provide a tool set that can be used in whole or in part.– If you just want GFac, then you can use it without
buying an entire framework.
• Outsource security, information services, data and metadata, etc to other providers.– MyProxy, TG IIS, Globus, Condor, XMC Cat, iRods, etc.
Advanced Support Scenarios
• GridChem/ParamChem workflow support• UltraScan Job Submission (GFAC)• EST Pipeline
– Bioinformatics pipeline for managing mass job submission.
More Information• This is downloadable, packaged software.
– Apache Maven build system provides everything you need to to build the gadget container, gadgets, workflow composer, and backing services.
– Get code by anonymous SVN checkout.• Email: [email protected],
[email protected], [email protected]
• OGCE Web Site: www.collab-ogce.org• Blog/News Feed:
http://collab-ogce.blogspot.com/
Acknowledgements and People
• Funding by TeraGrid GIG, RP and by OCI SDCI• IU: Marlon Pierce, Suresh Marru, Raminder
Singh, Archit Kulshrestha, Zhenhua Guo• TACC: Maytal Dahan, Rion Dooley• SDSC: Nancy Wilkins-Diehr, Jeff Sale• SDSU: Mary Thomas
GridChem Advanced Support
• IU Personal – Suresh Marru, Marlon Pierce, Raminder Singh, Ye Fan
• Gateways Area Director – Nancy Wilkins-Diehr• GridChem Team – Sudhakar Pamidigantam,
Rion Dooley, Vikram Gazula & Rest of GridChem team.
Empirical ForceFields Parameterization Need Process
Vanommeslaeghe et al. J. Comp.Chem 2010, 31, 671-690
Published by AAAS
A. J. Stone Science 321, 787 -789 (2008)
Fig. 1. Errors (V) in electrostatic potential on a surface at 1.8 times van der Waals radii around N-methyl propanamide for two models. (Left) Point charges; (right) charge, dipole, and quadrupole on C, N, and O; charge and dipole on H. The errors are much reduced in the multipole approach
Lack of Accurate Force Fields Produce Erroneous Property Estimation
Cyberenvironments for ParameterizationComputational Reference Data Generation
Molecular Force Field CyberenvironmentsParameter Initialization and optimization Workflow
Parameter definitions
Model/Reference Data Definition
Merit Function Specification
Consistency Checker
Optimization Methods Choice
Optmization Job Launcher
Update Parameter Database with new set
Workflow Manager
Optimization Incomplete?
Paramater testing Model
Successful Testing
Optimization Monitor
Optimization Job Completed?
Paramater Sensitivity Analysis
Notification of End of Workflow
Expert Interface
3131
OVP/RST/ MIG
OGCERe-engineer, Generalize,
Build, Test and Release
LEAD
OGCE Gateway Tool Adaption & Reuse
GridChem
TeraGridUser Portal
OGCE Team
GridChem
Ultrascan
BioVLab
ODI
Bio Drug Screen
EST Pipeline
Future Grid
GFac, XBaya, XRegistry, FTR
Eventing System
LEAD
Resource Discovery Service
GPIR, File Browser
Gadget Container, GTLab, Javascript Cog,
XRegistry Interface, Experiment Builder, Axis2 Gfac, Axis2 Eventing System,
Resource Prediction Service
Experiment Builder, XRegistry Interface
XBaya
Gfac, Eventing System
XBaya, GFac
Workflow Suite
Workflow Suite
Workflow Suite
???
Compute ResourcesCompute Resources
Resource Middleware
Resource Middleware Cloud Interfaces Grid Middleware SSH & Resource
Managers
Computational Clouds
Computational Grids
Gateway SoftwareGateway Software
User Interfaces
User Interfaces
Web/Gadget
Container
Web Enabled Desktop
Applications
User Managemen
t
Auditing & Reporting
Fault Tolerance
Application Abstractions
Workflow System
Information ServicesMonitoring
Registry Security
Provenance & Metadata Managemen
t
Local Resources
Web/Gadget
Interfaces
Gateway Abstraction Interfaces
Cyberinfrastructure Layers
Color Coding
Dependent resource provider components
Complimentary Gateway Components
OGCE Gateway Components
LEAD/OGCE Layered Workflow Architecture
Workflow Execution &
Control Engines
Workflow Execution &
Control Engines
Apache ODE
Workflow Specification
Workflow Specification
Workflow Interfaces (Design
& Definition)
Workflow Interfaces (Design
& Definition)
PythonBPEL 2.0
BPEL 1.0 Java Code Pegasus DAG
Scufl
XBaya GUI (Composition,
Deploying, Steering & Monitoring) Gadget Interface for
Input Binding
Condor DAGMan
Taverna
Dynamic Enactor
Jython InterpreterGBPEL
Flex/Web Composition
GFac Current & Future Features
Input Handlers
Input Handlers
Scheduling Interface
Scheduling Interface
AuditingAuditing
Monitoring InterfaceMonitoring Interface
Data Management AbstractionData Management Abstraction
Job ManagementAbstraction
Job ManagementAbstraction
Fault Tolerance Fault Tolerance
Output HandlersOutput Handlers
Registry InterfaceRegistry Interface
Checkpoint Support
Checkpoint Support
GlobusGlobus
Campus Resources
Campus Resources
UnicoreUnicore
CondorCondor
Amazon Eucalyptus
Amazon Eucalyptus
Color Coding
Planned/Requested Features
Existing Features
Gateway Computing Environments (GCE10)