SPECTRE A New Paradigm in Distributed Computing

advertisement
Distributed Image Reformation
Lockheed Martin
M&DS Reconnaissance Systems
Advanced SW Concepts Group
Advanced SW Concepts
Group
• Evaluate, prototype and integrate
newest software technologies into
LMM&DS-RS software systems.
• Primary focus
– Distributed Processing Systems
• Primary language
– Java
Advanced SW Concepts
Group
• Technology areas
– Remote Method Invocation (RMI)
– Jini
• JavaSpaces
– Rio
– JXTA
– Advanced Database Concepts
• Spatial databases
Image Reformation Problem
• Synthetic Aperture Radar data is
processed to look like a black & white
photograph
• This product is in a simple, but unique
proprietary format.
– Very similar to the RAW image format
• The produced image can be very large
– 2 GB not uncommon
Image Reformation Problem
• Our customers are wanting web-based
access to this data.
• Commercial web browsers do not
support the proprietary format
– They probably never will
– Size of product would make for a very
unpleasant download time.
Image Reformation Problem
• Solution
– Convert proprietary format to JPEG and /
or GIF image formats.
– GIF retains resolution
• Not much compression
– Still going to have very long download times
• BUT…you can look at it in a web browser
Image Reformation Problem
• JPEG
– Lossy compression
– Can be viewed in web browser
– If user is willing to accept some loss in
resolution, JPEG is an acceptable format.
Image Reformation
• Allow user to select GIF or JPEG file
reformation depending on resolution
required.
New Problem
• Due to the size of the imagery,
reformatting to GIF or JPEG is a time
consuming (and expensive)
undertaking.
• On a Sun 6500 machine
– Average 25 minutes for conversion to
JPEG
– Longer than it takes to produce the product
in the first place
Proposed Solution
• Reformation via Distributed Processing
Architecture.
• Dedicate a “network” of hardware to the
reformation task.
• Parse large original image into “chunks”
to be reformatted on various network
machines.
Enabling Technology
• Jini / JavaSpaces
– Java language (hardware independent)
services
• Object Flow Architecture
Architecture
• Parse original image into Java language
objects.
• Put those objects into a JavaSpace.
• Deploy reformatting Jini enabled
services to network machines.
• Services retrieve imagery objects from
JavaSpace, reformat, put objects back
into JavaSpace, repeat until done.
Architecture
Object Reformation
Service
Object Reformation
Service
Image Decompostion
Service
Object Reformation
Service
Raw image object
Object Reformation
Service
JavaSpace
All services reside on different hardware platforms
Architecture
Object Reformation
Service
Object Reformation
Service
Image Decompostion
Service
Object Reformation
Service
Raw image object
Object Reformation
Service
JavaSpace
Architecture
Object Reformation
Service
Object Reformation
Service
Image Decompostion
Service
Object Reformation
Service
Raw image object
Reformatted image object
Object Reformation
Service
JavaSpace
Architecture
Object Reformation
Service
Object Reformation
Service
Image Recompostion
Service
Object Reformation
Service
Raw image object
Reformatted image object
Object Reformation
Service
JavaSpace
Etc, etc, etc...
Design Concerns
• Optimal size for image objects.
– Network bandwidth, latency
• Book keeping system
– What chunk goes where?
• Jini / JavaSpace technology is very
new, it is evolving at a rapid pace.
– V1.1 is current
– V1.2 is in Release Candidate state
• has not been tested with Java 1.4
References
• Professional Jini
– Sing Li - Wrox Press
• JavaSpaces Principles, Patterns, and
Practice
– Freeman, Hupfer, Arnold - Addison Wesley
• Core Jini
– W. Keith Edwards - Prentice Hall
References
• www.jini.org
• www.javasoft.com
• subscribe to:
– JAVASPACES-USERS@JAVA.SUN.COM
Point of contact
• Bill A. Rawlings
– Applications SW Engineering Manager
– Advanced SW Concepts Group
– Lockheed Martin M&DS Reconnaissance
Systems
– bill.a.rawlings@lmco.com
– (623)925-7574
Download