Project ideas for the PRIME 2012 program
Scope of the projects: GAMESS toolbox
The Grid Computing Computing Competence Center (GC3) of the university of Zurich is developing a GAMESS toolbox: a series of add-on features to enhance the usability and reliability of GAMESS-US when used for large scale data analysis. The proposed projects are all contributing to the GAMESS toolbox.
All projects are centered around the GC3Pie execution framework and GAMESS-US
The candidate will learn the inner workings of GAMESS-US as well as how to develop utilities in the Python programming language.
As part of the project, the candidate will work closely with the GC3 group as well as with computational chemists from Kim Baldridge's research group.
If you find yourself interested in any of the proposed ideas, do not hesitate to contact us at info@gc3.lists.uzh.ch.
1. Automated GAMESS-US release validation suite:
For any new release of GAMESS-US a validation suite will automatically verify the consistency of the newly-released code against a defined reference data set. The validation suite will deploy the new release of GAMESS-US into a virtual appliance, compile it following a standard procedure, and run one instance of this new virtual appliance for each reference data set. The candidate will have to extend the already available GAMESS/GC3Pie execution framework to implement the verification of results.
Each of the provided data-sets will define the validation condition that the GAMESS-US execution will have to meet; part of the project consists in translating the test conditions (specified in plain English) into a form that can be read and tested by a computer (e.g., Python code, or some form of a configuration file).
By default the GAMESS-US standard examples will be used. The suite will also allow to define an alternative data-set provided it is predictable (and thus verifiable) the behavior of GAMESS-US.
2. GAMESS-US restart mechanism
Most of the time running GAMESS-US requires a tightly coupled system with lot of resources available at the same time and for an extended period of time. Local institutions access policies often limit execution time and/or the number of concurrent processes. This constrains considerably the use of GAMESS-US. However, GAMESS offers a possibility to resume an interrupted GAMESS-US execution by extracting from a partial output file the last consistent registered operation. This project aims building an automated mechanism to retrieve the partial output of an interrupted GAMESS-US execution and to resubmit it as its continuation. This mechanism will have to be transparent for the end-user.
3. MyGAMESS portal
The project aims building a web portal to ease the management of large analysis campaigns, to keep molecules as well as results organized in a project-like structure. The candidate will have to develop a web portal using "off the shelf" web technologies (names that come to mind are Google Closure, Twitter Bootstrap, and Django), integrate the GC3Pie execution engine to allow an efficient and controlled execution of GAMESS-US, develop a User Interface that would be centered around the molecule organization model.
The resulting web interface will allow researchers to organize their computational activities and results; automatically schedule GAMESS-US analysis, extract aggregated information from the available results. Ultimately the portal will integrated graphical editors and visualizers for the input/output data. The candidate(s) will have several interactions with computational chemistry users to learn and understand usage models and usage patterns.
This is a potentially very large project, that cannot be completed in one single student project. What we expect is to bootstrap the project and have a portal with simple basic functionality implemented.
4. GAMESS-US in MPI-enabled virtual appliance
As part of the support for GAMESS-US on our Swiss national grid infrastructure, we allow running GAMESS-Us within virtual appliances, through the AppPot mechanism. At the moment, AppPot allows GAMESS-US to be executed only in single-core mode. The project aims to extend AppPot to support MPI-based parallel execution of GAMESS-US.
5. GAMESS-US benchmark
A variant of project 1. is focusing on benchmarking of research results against standard databases available in the community. For each new release of GAMESS-US (or for each new method implemented), the project will allow to verify it against an on-line data base to check research results and performance.