Home -> Research Groups -> Cooperative Computing -> Projects -> Distributed Marine Environment Forecast System

Project
Distributed Marine Environment Forecast System
Grant Number N00014-00-1-0886.


In Collaboration with:
Office of Naval Research

PI: Tomasz Haupt
Team Members: Sheikh Ghafoor, Anand Kalyanasundaram


Description:

The world's geopolitical structure has changed radically in the past decade. Breathtaking changes in global and regional alliances have fundamentally restructured the nature and mission of America's defense forces. The overall mission of the U.S. Navy remains unchanged, but to keep open lines of global trade, the places where the Navy will likely be called on to do its duty have changed dramatically. No longer merely called on to maintain open-ocean superiority, today's Naval operations are increasingly littoral and regional. Oceanic prediction in the littoral is vastly more complex than in the open ocean. The temporal and special scales of variability are much shorter; the parameters of DoD interest, such as waves, storm surge, optical clarity, tide, sediment transport, beach traffic ability, currents, temperature, salinity, etc., are very different. In addition, the real time or even historic observations of these parameters, in many areas of potential interest, are either very restricted or limited at best. This rapid shift in emphasis has left the computational and modeling wing of the U.S. Navy with an emerging need to address this increasingly important style and scale of operations.

The ability to simulate the littoral environment rapidly, accurately, and across many temporal and spatial scales poses very significant challenges to gathering, processing, and disseminating information. Coupled with a military that must move and respond quickly in order to maintain a state of information superiority, computational demands have risen dramatically in the past ten years. At the same time, underlying computer architectures have undergone an equally drastic change. No longer in the business of making vector supercomputers, U.S. computer manufacturers produce multiprocessor computers made from commodity processors with a high-speed interconnect. While such architectures offer vast improvements in cost versus performance, software written for
vector supercomputers must be extensively rewritten to run on these new architectures. Thus, there is an emerging demand for scalable parallel models that not only can run on current architectures, but also can be easily adapted to new ones. Finally, the exponential growth in data communications, exemplified by the Internet infrastructure and web applications, are enabling dramatic advances in distributed computational environments.

DMEFS is an attempt to integrate the expertise of computer scientists with METOC modelers and applications developers to produce the ITI that can expedite the progress in the field of Ocean Science. Exploiting leading edge developments in computational grid, web, and component software technologies, DMEFS integrates some of their best features to provide a multi-tiered infrastructure that links a user on a desk top via a network application to model execution on high performance computational platforms that can be distributed globally. The services that can be provided by the DMEFS infrastructure are tailored to minimize the complexity of model research, development, transition, and operation or execution.

DMEFS ITI: Enterprise Computational System

DMEFS is a Grid Portal
The DMEFS user located anywhere (given the Internet access) may perform operations using computers at four independently managed sites: ERDC MSRC, NAVO MSRC, MSU main campus and MSU/ERC extension at Stennis Space Center.

Security
DMEFS servers (Web Services, WSDL) are located at MSU main campus, and are accessed through SOAP over https. Positive authentication to the DMEFS server does not provide access to computational resources at MSRC or MSU. A valid Kerberos ticket (TGT generated using the standard kinit command in conjunction with SecurID) and valid (proxy) Globus digital certificate, respectively, are necessary to perform any operations on these resources. In each case the resource owner preserves the full control over authentication and authorization process.

Functionality
The DMEFS Grid Portal allows the user to access seamlessly remote file systems, monitor status of the remote compute servers, execute commands on the remote machine, submit and monitor batch jobs as well as complex, multistep applications. In addition, the DMEFS system provides persistency: configurations and results of all user tasks are stored in the database. The persistence also simplifies transition from R&D to operations.

Graphical User Interfaces guide the user in configuring the application: setting the values of model dependent parameters, location of the input files, target machine, etc., providing default values for all parameters. Optionally, the user may specify name of a batch queue. In such case the DMEFS will automatically generate a batch script (PBS, NSF or Load leveler), seamlessly moving files across file systems, as needed.



METOC applications in DMEFS


Learn More




Status of the Project
A prototype of the system has been demonstrated at DMEFS Workshop
at Stennis Space Center, March 18-19, 2003,
and Oceanology International Conference - Americas 2003
in New Orleans, June 3-6, 2003.
The slides from these demonstrations are available from: