05.06.2013 Views

PNNL-13501 - Pacific Northwest National Laboratory

PNNL-13501 - Pacific Northwest National Laboratory

PNNL-13501 - Pacific Northwest National Laboratory

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

• securely moves files to a target resource and submits<br />

job<br />

• provides reporting on progress of job and allows jobs<br />

to be terminated<br />

• monitors the running job, parsing text output files and<br />

returns data to the client<br />

• makes use of existing infrastructure thus providing<br />

for low-cost deployment.<br />

Distributed Data Management<br />

We developed a set of written requirements for distributed<br />

data management through our interactions with scientists<br />

and engineers in chemistry, atmospheric sciences,<br />

manufacturing, materials, and electronic notebooks.<br />

Alternative technologies for addressing these<br />

requirements were investigated. We focused our efforts<br />

on prototyping solutions with the distributed authoring<br />

and versioning protocol, a proposed internet engineering<br />

task force standard. We installed an Apache web server<br />

with a plug-in module for providing back end data and<br />

meta-data storage and performed load and stress testing.<br />

Our research included the development of a three-layer<br />

client application programmer interface. This layering<br />

technique provides clean separation of the protocol<br />

implementation, a data storage abstraction, and an object<br />

abstraction. We also developed the concept of the virtual<br />

document as a key abstraction for dealing with complex<br />

concepts such as calculations or workflows.<br />

Resource Information Services<br />

The primary function of a resource information service is<br />

to provide yellow- and white-page access to information<br />

about system- and user-defined resources. The<br />

incorporation of a resource information service into the<br />

computing infrastructure provides for the use of common<br />

data representations and a shared data source across<br />

projects and organizations.<br />

We researched the use of meta-languages to support the<br />

definition and manipulation of objects in an resource<br />

information service. We settled on the use of the<br />

developing XML Schema standard as the meta-language<br />

and adopted and extended the Directory Services Markup<br />

Language, another developing standard, as a dialect.<br />

Using this dialect, we defined resource information<br />

objects based on previous investigations and our<br />

continuing work with the Gridforum Grid Information<br />

Services working group. We developed software to<br />

generate Netscape Directory Server schema and a Java<br />

binding for accessing and manipulating the objects from<br />

the XML-based object definitions. The use of a metalanguage<br />

and automatic code generation provides an easyto-use<br />

application programmer interface for collaborative<br />

problem-solving environments developers, an easy way to<br />

modify and extend the object definitions, and an<br />

abstraction that allows for the replacement of the<br />

underlying service without affecting the applications that<br />

rely on the service. Significant effort was invested to<br />

develop a layered application programmer interface that<br />

hides the implementation details of the underlying service<br />

and communications protocol.<br />

Atmospheric Sciences Prototype<br />

We worked closely with <strong>Laboratory</strong> atmospheric science<br />

researchers in many analysis and design sessions. The<br />

purpose of the collaboration was to gather user<br />

requirements for a collaborative problem-solving<br />

environment for regional climate change modeling and to<br />

develop a conceptual prototype. The resulting prototype<br />

provides an overall experimental process representation or<br />

workflow for doing atmospheric research and coupling<br />

impact assessment models for researchers and policy<br />

planners. The prototype is being used to drive an<br />

understanding of the requirements for developing a new<br />

software suite for atmospheric sciences.<br />

Summary and Conclusions<br />

The application of web-based technologies to problemsolving<br />

environments allows the scientific community to<br />

benefit from the robust growth of low-cost commercial<br />

and open-source technologies. We successfully applied<br />

these technologies in the development of multiple<br />

components. We found the performance of http to be<br />

viable for expected use patterns and the prospects for<br />

sufficient scalability to be within reach for many<br />

applications.<br />

We concluded that discovering key abstractions is very<br />

important to component-based architectures for two<br />

reasons:<br />

1. it insulates the client program from changes to the<br />

underlying technology<br />

2. it enables the insertion of a middle layer that can<br />

provide service mappings or translations thus<br />

enabling interoperability.<br />

Additional research is required in this area.<br />

Computational Science and Engineering 107

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!