zzz

1
CIPRES AToL Development and collaboration
environments
Mark A. Miller
Principal Investigator, Biology San Diego
Supercomputer Center
2
CIPRES Libraries enable fine grained
communication between programs
3
CIPRES Libraries were exposed to the community
through the CIPRES Portal
4
CIPRES the CIPRES Portal parses input files, and
provides users with appropriate tool selections.
5
Features

• PAUP
• RAxML (bootstrapping)
• GARLI
• MrBayes
• ClustalW
• RecIDCM3 boosting for PAUP and RAxML
• ReST accessible
• Supports Nexus, Phylip, and Hennig86 formats.

6
Usage statistics for all jobs
7
Usage statistics by Job
8
Problem

• CORBA architecture has a very high overhead for
  adding new data types/services
• This makes it hard to expose to users
  capabilities that are already available on the
  command line.

9
Solution

• Create a new architecture that is designed for a
  portal, rather than for integrating to existing
  tools.
• The application can still have knowledge about
  the data, but the data is treated as text
  strings, and not as CORBA objects.
• This decreases the overhead a lot.

10
Presentation layer Connects to SDK by 2 Java
Classes
Presentation layer Connects to SDK by 2 Java
Classes
Generic Architecture For handling Queries And
Tasks
Central Registry defines Resources and data
types.
Support for access to Many kinds of
remote resources
11
www.ngbw.org
Generic Architecture has been implemented for the
Next Generation Biology Work Bench. It allows
users to store their jobs locally.
We mounted 50 tools in 80 days using this
architecture
12
www.ngbw.org
Interface allows tool selection and
configuration
13
All command line parameters can be set.
www.ngbw.org
14
Here is where we are headed..
Public Data Providers
Local Data Storage
TeraGrid (group allocation)
IBMGoogle
AToL Resource Storage
NSF Allocated Data Storage
AToL Framework
Google Data Storage
HPC allocations (NRAC, etc)
Local User Resources
SRB/Grid Bricks
TeraGrid (individual allocation)
SDSC fee-for-service DataBase
SDSC Resources
The generic architecture allows users to easily
add data and computational resources.