| Abstract: | Introduction
In recent years, the United States Geological Survey‘s (USGS) National Mapping Discipline (NMD) has expanded its scientific and research activities. Work is being conducted in areas such as emergency response research, scientific visualization, urban prediction, and other simulation activities. Custom-produced digital data have become essential for these types of activities. High-resolution, remotely sensed datasets are also seeing increased use.
Unfortunately, the NMD is also finding that it lacks the resources required to perform some of these activities. Many of these projects require large amounts of computer processing resources. Complex urban-prediction simulations, for example, involve large amounts of processor-intensive calculations on large amounts of input data.
This project was undertaken to learn and understand the concepts of distributed processing. Experience was needed in developing these types of applications. The idea was that this type of technology could significantly aid the needs of the NMD scientific and research programs. Porting a numerically intensive application currently being used by an NMD science program to run in a distributed fashion would demonstrate the usefulness of this technology.
There are several benefits that this type of technology can bring to the USGS‘s research programs. Projects can be performed that were previously impossible due to a lack of computing resources. Other projects can be performed on a larger scale than previously possible. For example, distributed processing can enable urban dynamics research to perform simulations on larger areas without making huge sacrifices in resolution. The processing can also be done in a more reasonable amount of time than with traditional single-threaded methods (a scaled version of Chester County, Pennsylvania, took about fifty days to finish its first calibration phase with a single-threaded program).
This paper has several goals regarding distributed processing technology. It will describe the benefits of the technology. Real data about a distributed application will be presented as an example of the benefits that this technology can bring to USGS scientific programs. Finally, some of the issues with distributed processing that relate to USGS work will be discussed. |
| Genre: | USGS Numbered Series |
| ProdID: | 50597 |
| Citation Author: | Maddox, Brian G. |
| Citation Contributing Office: | Geological Survey (U.S.) |
| Citation Datum: | |
| Citation Day: | |
| Citation Edition: | - |
| Citation Editor: | |
| Citation End Page: | |
| Citation Issue: | |
| Citation Keywords: | |
| Citation Language: | ENGLISH |
| Citation Larger Work Title: | |
| Citation LatN: | |
| Citation LatS: | |
| Citation LonE: | |
| Citation LonW: | |
| Citation Month: | |
| Citation No Pagination: | N |
| Citation Number Of Pages: | |
| Citation Online Only Flag: | |
| Citation Phsyical Description: | 16 p. |
| Citation Projection: | |
| Citation Public Comments: | |
| Citation Publisher: | Geological Survey (U.S.) |
| Citation Series: | Open-File Report |
| Citation Series Code: | OFR |
| Citation Series Number: | 2002-15 |
| Citation Search Results Text: | Beowulf Distributed Processing and the United States Geological Survey; 2002; OFR; 2002-15; Maddox, Brian G. |
| Citation Start Page: | |
| Citation Volume: | |
| Citation Year: | 2002 |
| Type: | citation/reference |
| Text: | Beowulf Distributed Processing and the United States Geological Survey; 2002; OFR; 2002-15; Maddox, Brian G. |
| URL (THUMBNAIL): | http://pubs.er.usgs.gov/thumbnails/usgs_thumb.jpg |
| URL (INDEX PAGE): | http://pubs.usgs.gov/of/2002/0015/ |
| Date Other: | Sat, 1 Jan 1994 00:00 -0600 |
| Publisher: | Geological Survey (U.S.) |
