SALICIN: a data-intensive 

Real-time access to large scientific and heterogeneous environmental data has
prompted the need for the development of information servers that integrate
archival-storage, database, world wide web user-interface, and
visualization technologies.

A project called SALICIN (anagram for NALCIIS, North American Landscape
Characterization --NALC-- Internet Image Server), developed by researchers
from the San Diego Supercomputer Center (Richard Marciano) and the 
National Supercomputing Center for Energy and the Environment at the 
University of Nevada Las Vegas (Bahram Nassersharif, Ringo Ling) is an example
of a distributed data-intensive application where remote sensing data sets
can be browsed, queried, processed, and downloaded over the Internet.

This repository consists of 60-meter resolution Landsat satellite multispectral
data sets (MSS data), as well as associated normalized 
differential vegetation index data sets (NDVI data), and digital elevation
data sets (DEM data), covering the entire north american continent and Mexico.
The data sets are in triplicate format, spanning three decades (70s, 80s, 90s),
thus, allowing the study of human influences and changes on the environment
and ecology over a thirty year period.

The SALICIN system attempts to integrate "off-the-shelf" high performance 
computing system and software architectures in order to: (1) develop a useful
working system based on current technology, (2) disseminate the
NALC data sets to the research community (in particular for global warming
studies), (3) develop a scalable working methodology for the future design of 
data-intensive applications based on upcoming petabyte/teraflop systems.

The storage requirements of SALICIN are roughly 0.25 Terabytes of 
satellite data, consisting of 5000 x 5000 pixel band data.  The system 
architecture comprises four layers:

    (1) World Wide Web User Interface
    (2) Subsampled Image Browser
    (3) Metadata Management Subsystem
    (4) Mass Storage Image Server Subsystem.

Because the typical speeds of today's networks and display systems lack the
necessary performance to handle real-time browsing of large satellite images,
smaller versions ("thumbnail" version) of the actual images must be used for
browsing and image selection.  Layer (2) deals with the storage and serving
of subsampled images (typically 200 x 200 pixel images) that can be inlined
on the fly into HTML documents served by a WWW server, and browsed through
Mosaic or Netscape type clients.

Layer (3) implements a centralized metadata database using the Oracle RDBMS 
allowing researchers to access the actual satellite data by content or 
metadata attributes.

Layer (4) makes use of available terabyte mass storage devices to store the
full-sized satellite data.

SALICIN is a highly distributed application in the sense that each layer
can reside on a separate platform.  The current prototype implements layer (2)
using CGI web scripts written with the Perl language and served by a Convex 
supercomputer, layer (3) using Oracle and Pro*C on a Sun Sparc server, 
and layer (4) on a Cray supercomputer which controls a StorageTek Robotic silo 
of capacity 2.7 Terabytes.  The project integrates web user-interface, 
relational database, and archival storage technologies by a networking 
programming layer (socket interface), allowing web-based queries to be built
through the  standard Form mechanism to be translated to metadata queries
that search and retrieve massive data sets.  Issues of bandwidth and data
compression are also examined.

The SALICIN repository system is meant to serve as a platform on top of which
"value-added" data-intensive applications can be developed such as neural-net
classification tools for Landsat image classification, computational learning
laboratories for environmental curricula, and data mining of 
environmental changes across decades.  

SALICIN explores some of the design issues of the systems to come which 
will unlock the massive amounts of environmental data waiting to be mined and 
fully utilized.