Efficient Use of Networks for Browsing Image Data

George Milkowski 
James Gallagher

Graduate School of Oceanography
University of Rhode Island
Narragansett, RI 02882-1197

george@zeno.gso.uri.edu
jimg@dcz.gso.uri.edu

Image browsing is an important part of research activities involving satellite
data.  Supporting this function over a wide area network requires a
substantial and on-going commitment of time and resources on the part of the
sites providing this type of service.  As a result, browse service providers
are concerned about developing methods for providing browse data that are
efficient and effective for users.  In this paper we evaluate the performance
of different client-server systems accessing imagery data in order to determine
what factors influence their effectiveness and network through-put.

A comparison of two different client-server implementations that access the
same satellite image archive was made to determine differences in their user
statistics and the loads that they placed on network resources.  The two
client-server systems are: XBrowse, an application specifically designed for
browsing satellite sea surface temperature imagery over a network, and the
URI SST Archive WWW page an html forms based implementation.  The systems
access an image archive of 20,000+ processed sea surface temperature images.
The archival data is 1024x1024x8bit satellite imagery that has been processed
to sea surface temperature and remapped to a region approximately 60N-9N and
96W-34W at a nominal 5km/pixel resolution.  The period of time that the
satellite data covers is April 1979 to the present.  The usage statistics of
XBrowse and the WWW URI SST image servers are based on remote access session
logs of the two systems from between September 1994 to March 1995.

Analysis of the server usage statistics for the two systems indicates that
both systems provide effective access to the image archive based on the total
number of images transferred during the survey period and the average number
of images transferred per day.  However, there are significant differences
between the systems in terms of how efficiently image browsing is performed
based on the average number of bytes/image browsed XBrowse users required 1/3
the bandwidth of WWW users to perform their browse activities.  In addition
while the number of discrete users who accessed the image archive using the
WWW interface was a factor of 10 times larger than those that used XBrowse,
the total numbers of images access by XBrowse users is almost 2 times greater
than by the WWW server.  The results of this study highlight some of the
issues related to providing imagery data on-line and also provides some
interesting insights on user usage patterns.