Computational Chemistry Grid

1
Computational Chemistry Grid the Chemistry
Community
Kent Milfeld milfeld_at_tacc.utexas.edu
June 12, 2006
Supported by the NSF NMI Program under Award
04-38312
http//www.GridChem.org
2
Computational Chemistry Grid Overview

• Computational Chemistry Grid (CCG)
• What is GridChem? (Demo)
• Community The Service Challenge
• Web Services Resource Framework (WSRF)
• Handling Community Consulting and Allocations

What can you take home with you.
3
Computational Chemistry Grid Overview

• Infrastructure for Running Chemistry Applications
  through a Community Organization
• 3-year NSF-funded project
• 5 sites

4
Computational Chemistry Grid Overview
NCSA
CCS
OSC
CCT
TACC
5
Computational Chemistry Grid Overview

• NMI funded integration not development
• Goal cyberinfrastructure for
  computational chemistry community
• Run QM Chemistry Applications at HPC Sites
• Use GridChem Client
• Middleware with Grid Services

6
Computation Chemistry Grid? VO

• Grid-enabled Distributed Resources to routinely
  run chemical physics applications
• Desktop Environment integrated into an
  infrastructure for a specific community of users
• comp. chemists with large/small scale needs
• experimental chemists who occasionally need
  simulation capabilities to verify experimental
  results
• ?Virtual Organization-- establishes a distributed
  infrastructure for open scientific research

7
Computational Chemistry Grid Hardware
8
Computational Chemistry Grid Management

• Client Development Team
• App Interfaces, Visualization, Inputs
• Middleware Team
• CGI / Web Services
• Database Team
• DB design / implementation /maintenance
• Applications / Job Launchers

• EOT
• Training
• University Course modules
• Community Management
• Allocations
• Consulting
• Site Monitoring

9
GridChem

• Desktop Client
• Java (portable)
• Web Start Updates (Easily Installed Updated)
• Runs HPC Applications
• Job Submission, Deletion
• Job Monitoring
• Job Management

10
GridChem

• Chemistry Environment
• Build Molecules
• Construct QM Inputs with consistent
  commands/options
• File Browsing (Mass Storage)
• Authentication
• Output Parsing
• Visualization

11
Computational Chemistry Resource

• Over 400 processors and 3,525,000 CPU hours
  available annually

12
Applications

• Applications
• Gaussian
• Gamess
• Molpro
• NWChem

Compute Intensive 0.1-100 hrs runs
High Speed Infrastructure Not Needed
Batch Support
Secure Access (grid-enabled)
4 16 CPUS/job
13
Demo
14
Community Challenges

• Community Allocations
• CCG User Map to Site Community Account
• Multi-site Job Accounting
• Applications ? Site Interfaces
• Site, Client, and Services Monitoring
• Migrating Client Services to New Software
• Consulting, Problem Reporting
• File Name Space

15
Old GMS Services (simple view)
ssh
ssh
16
Putting WS into GMS?GMS_WS (simple view)
Author Rion Dooley _at_TACC
Serialized XML Xstream
Web Application Server
Grid API
Grid API
17
Demo
CGI Services
18
WS Services
WS Services
19
Demo
WS Services
20
Demo
WS Services
21
Client ??Objects ?? Database Interaction
DTO
Client
Business Model
WS Resources
DTO (Data Transfer Object) Serialize transfer
through XML DAO (Data Access Object)
How to get the DB objects hb.xml (Hibernate
Data Map) describes obj/column data mapping
DAO
Objects
Hibernate
Database
hb.xml
22
Database Table Relationships
Jobs
userID projectID resourceID loginName SUsLocalUser
Used
jobID jobName userID projID softID cost
Resources
resoruceID Type hostName IPAddress siteID
23
Computational Chemistry Resource
24
Consulting

• Exclusive requires login
• Easy to Use web based
• Ticket Bases MySQL database
• Low Maintenance direct updates
• Reasonably Smart designed for HPC
  preferences definable view searchable
  by consultant
• Supported by TACC, version 2.0.1

25
Consulting
www.gridchem.org/consult
26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
(No Transcript)
31
(No Transcript)
32
(No Transcript)
33
Allocations ?Project

• Allocations
• Peer reviewed (committee for 1000 SUs)
• Accounting
• Job based
• Project Oriented (PI Control, Community/External)
• Add Users
• Set Limits on Research Group Members (capable).

34
Requesting Allocation

• Registers as a GridChem User
• User Name / Password / BIO ? database
• Allows access immediately to consulting ---
• Password-Protected FORMS
• Community Request (community SUs)
• External Project Request (externally allocd SUs)
• Approval
• Allocation Reviews nothing remotely like TG
  POPS
• Administration DB interface makes it easy.

35
Personal Information
36
User Name Information
37
Personal Information
38
External Allocations
39
Summary

• Client Access to HPC Systems is reasonable and my
  become popular.
• Communities Require Organization beyond what HPC
  Resources can provide.
• A Web Service Frameworks (WSResourceF) makes
  sense.
• Both HPC Sites Communities can agree upon
  services in an organized way.