| Title of Project: | Distribution of the Decision Evaluator for the Cattle Industry across the WWW | |
| Project Director: | Christopher J. Kaiser | |
| Applicant Organization: | University of Missouri at Columbia |
This project will increase the access to technology in two ways. First, using any common web browser, people will be able to run the program by submitting data to a secure web site. This site will be served from a high speed computer cluster located in the Animal Sciences Unit. Second, the software will be made available to the public for installation on their own server or redistribution under a general public license. This license would permit researchers to obtain the source code and incorporate their results into the model, or validate it under unique production conditions.
| Congressional District Number: | ninth | |
| Period of Proposed Project Dates: | 09/01/1999 to 08/31/2000 |
First, creating a web based version of the Decision Evaluator for the Cattle Industry or DECI, a production simulation model, will allow us to distribute it across networks at minimal to no cost. Thus, we will be able to manage the program on a single server, concentrating the need for computer expertise and resources, while providing the widest access possible. Computer labs allowing student access to the internet are readilly available on college campuses. In addition, most extension offices and schools, as well as many homes now have, or soon will have, internet connections. Web server clients, including Netscape Communicator and MS Internet Explorer, are available for free. In some cases, it would not even be necessary to own a computer. With this program, it would be possible for an off campus student to access a high powered server using a television top web browser costing less than $100, or a public access computer at a local library.
This project will also expand research and extension efforts in production systems modelling. The DECI program will be written in java, which is designed to be a write once run anywhere environment. That is, by using a java virtual machine, software that converts java binaries into machine specific code, the same software can be run on virtually any architecture or operating system. Researchers will be able to more easily collaborate on future refinements to the underlying model without worrying about compatibility issues. Moreover, this project is part of an overall effort of researchers at several institutions to develop DECI under an open source license. We see this as a prototype for collaboration between scientists on developing both research and educational/extension software.
This project will most directly address the secondary goal of training students for careers in agriculture and related fields. Initially, we will use the software as part of the beef production and management course in this department. Once proven, we will make it available to other beef production instructors at other colleges and vocational agriculture instructors at high schools. This project will also benefit the universitys distance educatiton efforts because the program will be distributed across the internet, allowing us to potentially offer beef production to off campus students.
This software will allow realistic simulation of common production practices, so that students can evaluate effects of management decisions and gain an understanding of necessary resources and practices. For example, by making repeated runs, they will be able to evaluate "what if" scenarios, to see the effect of using different types of cattle, feeding more or less supplemental feeds, or calving at different times of the year.
The design of this computer is based on that of the Beowulf project (http://www.beowulf.org), which uses off the shelf commodity hardware and open source software to build high speed parallel machines at a minimum cost. Several of the larger clusters have been ranked among the 500 fastest computers in the world.
This particular server is currently being utilized for computationally intensive genetic analyses and production systems simulations. It is designed to have high availabilty and be able to run a large series of analyses rapidly. This computer has typically gone over 200 days without being rebooted, including system upgrades.
The DECI program has developed over several stages. At its core is a stochastic animal model based on the interaction of many biological mechanisms, among these are feed intake, energy metabolism, growth, and reproduction. While very powerful and flexible, working with the original model required editing a series of text files or even modifying source code and recompiling. To make the model more accessible, a graphical user interface running on MSWindows platforms was written. Simulation results are then evaluated using Standardized Performance Analysis guidelines. Designed as a decision support tool for cow-calf producers, the software has also been used as an instructional aid in university beef production courses. However, this version of the program has several shortcomings which we seek to address with this project.
Installing and running the program on local machines has been problematic at best with installations failing on many machines and requiring additional libraries to be downloaded on others. This difficulty is eliminated by running it from a dedicated web server. Users would submit data and receive back results via cgi web pages, a task with which virtually all computer users are very familiar. Moreover, serving the program over the web would enhance distance education efforts by allowing students access to course material from any location. A second concern with the current model is the required compute time. The computer cluster on which we would host the web version would be able to run multiple copies in less than one third the time required to run it on typical Windows desk top machines. Finally, in order to gain the simplicity in the current front end, many capabilities of the base model were removed or limited. The new front end would return these functions to the model.
By serving a more versatile version of the DECI program across the web, this project meets the overall objective of ADEC, to make optimal use of available resources for agricultural extension, resident education, and research by sharing resources between participating institutions. Specifically, it meets the third objective of training students for careers in agriculture, natural resource management, environmental science, human sciences and the food industries.
This project also addresses other goals of ADEC. The DECI model is licensed under a general public license or GPL. Under the GPL, researchers will be able to obtain the source code to the model and modify it in any way. In addition, these modifications will also be freely available. Scientists would be able to incorporate their research into the model and evaluate it from a systems based perspective. This will facilitate interaction among leading agricultural scientists, the fourth objective.
By providing a means for producers to use the latest research to evaluate their management system, this project also addresses ADECs second goal, to improve the competitive position of United States agriculture in international markets by disseminating information to producers, processors, and researchers.
Because it is designed as a decision support tool, researchers and individual producers can use DECI to assess the economic costs and benefits associated with alternative management practices, including marketing animals which meet branded product or niche market specifications, use of new or alternative feed crops, and managing their herd to reduce adverse environmental impacts. Thus the project meets the fifth objective of enhancing the ability of United States agriculture to respond to environmental and food safety concerns, and the sixth objective of identifying new uses for farm commodities and increase the demand for United States agricultural products in both domestic and foreign markets.
The initial design phase will entail determining necessary information for the model, the layout of input screens and presentation formats for results. While incorporating some features of the Windows version of DECI, we will seek to maximize flexibility and allow the user to take advantage of as many features of the underlying model as possible.
Once the interface layout is determined we will develop a beta version for use in the beef production and management course. We will ask students to evaluate the program, focusing on ease of use, clarity of layout, and performance of the model.
This feedback will be used to refine the interface, and develop on line instructions. Following this, we will seek other beef production instructors willing to evaluate the model and make it available to them on our server.
Using their input we will make final refinements to the model and release a stable version.
In addition to making the program available on our web server, we will also make it available for downloading. This release will be done under a General Public License. The terms of this license will encourage other researchers to contribute to further software development, by making the source code to the program freely available to them. The only restriction on use of the software is that changes to the code are also made available under the same terms.
This program could also be made available to cattle producers with small to medium sized operations. These producers have had limited access to management expertise in the past.
Finally, because the software is open source, it will be available to other researchers who can not only use the model to evaluate production practices but also improve and update the model by incorporating the latest biological and economic research.
These efforts include converting the underlying simulation software to an object oriented structure, making it easier for collaborators to incorporate new research models; updating some aspects of the models biology; validating the model in a wider variety of production conditions; incorporation of one or more mechanistic forage simulation models, and development of an extension program based on DECI.
The majority of this work is currently being done at Colorado State University, where the model was originally developed, and at the USDA Meat Animal Research Center, where the original graphical interface was written and the model was validated against data from the Germ Plasm Evaluation project. Major contributors to the model development include Rick Bourdon, CSU and Charles Williams and Tom Jenkins, USDA-MARC.
The extension project will be funded by a development proposal that William Herring and I are submitting.
This project along, with the other efforts mentioned above, will serve as a demonstration of collaborative software development under an open source model. Our efforts will encourage further development, not only of this model, but of other similar projects as well.
The program will be distributed in two ways. First, it will be available from our web server for remote administration. Second source code for the entire program will be made available for download under the terms of the General Public License.
Making the software more widely available will serve as a catalyst to further model development by encouraging more researchers to contribute new features or enhancements.
Successful use of an open source development model for educational and research software will also encourage development of additional products by other individuals.
Finally, by creating a new means of distributing eductional software, this project will facilitate a more inclusive educational system. Traditional educational systems include either real or perceived barriers to under represented groups, by definition. By using a distributed computing model, access to the software is not bound to these traditional systems. Because the software requires less expensive hardware to run, and is available under a public license, economically disadvantaged people and organizations will have improved access to the latest technology.