Webinar 1
From Jini.org
Contents |
Webinar #1 - Distributed Image Reformation Using JavaSpaces Technology
Tuesday, September 28, 2004
Bill Rawlings - Software Engineer, Principal, SW Division Architect - Lockheed Martin Integrated Systems & Solutions, Lockheed Martin
Abstract
More customers are requesting information via web-based services. Since the predominant browsers only support a few image types, the reformation of an image from one type to another can be a time and resource consuming process.
Lockheed Martin Integrated Systems and Solutions has developed a system based on Jini and JavaSpaces technology which significantly reduces the time and resources required to reformat an image. Using JavaSpaces and the classic Replicated-Worker pattern we have been able to convert 10k x 10k raw imagery into .JPEG format in under 5 minutes on a cluster of Apple based G4s. This task averaged 30 minutes on a single CPU on a Sun 6500 class machine.
There are many design considerations and trades involved in successfully designing and deploying such a system. Network and disk I/O limitations heavily influence the optimal configuration of the JavaSpace based distributed system, and a considerable amount of testing is required to arrive at the optimal solution.
This presentation will describe the system (called "Alien Race") and its architecture. It will also discuss testing and optimization techniques we used in its development, the replicated-worker architecture and benchmark results obtained from extensive testing.
In summary, this presentation describes a simple and cost-efficient method to reformat large amounts of imagery data. The system described could be of benefit to any person or company dealing with imagery on a large basis. Additionally, the concepts presented in this paper should be of interest to anyone utilizing or considering the utilization of Jini/JavaSpaces technology.
