heavily on the underlying architecture.
ery software. Backups are generally faster and more
stable. Recovery from disasters and hardware failure
Problem: Network operating speed
is faster and easier. It allows for a buildup of storage
devices in one location where they can all be moni-
Description
It was found that 10-Mb networks are no longer
tored easily and efficiently.
fast enough to serve the needs of most medium- to
With 100-Mb network architecture, it was found
large-sized offices. DLT drives do not operate effi-
that the options for data storage and manipulation
ciently at 10-Mb network speeds; file transfer speed
are considerably more flexible. With a properly con-
hampers access to data and backups.
figured 100-Mb network, access to data on remote
resources for manipulation appeared to be nearly as
fast as access to locally stored information.
Solutions
Sites should implement a 100-Mb network as
Local workstation computers may connect to the
soon as feasible. Management at all levels should
network hosts to retrieve and work with data. On
realize that if we don't upgrade our network capabili-
some sites, data managers have instituted software to
ties, our geospatial capabilities will be severely hand-
facilitate a centralized data repository with controlled
icapped.
access. Data is stored in the central database and can
It has been found, however, that once workstation
be "checked out" to only one user at a time. Thus
segments are upgraded to 100 BaseT, the 100 BaseT
coordination is enforced, eliminating creation of mul-
segment to the GD&S Server becomes overloaded.
tiple versions of one data file.
Three-tier networking can overcome this problem:
At some sites data creators and manipulators pre-
100 BaseT for clients, Gigabit or ATM for servers, and
fer to keep the data on their local PC while they are
a storage area network for backups. Two diagrams
working on it rather than in a shared database that
showing sample network architectures are available
they would have to access across a network. This may
in Appendix G.
be due to slow networking, unfamiliar downloading
interfaces, or just a personal preference to have pos-
Problem: Centralized vs. distributed storage and
session of the data file while working on it.
manipulation of data
Our study showed that in some cases version con-
trol through informal communications between those
Description
Problems can arise when two people have copies of
working on the same data was very well coordinated.
the same dataset on their own PCs and make changes
This works well for organizations with a high degree
in the data. The data may have been downloaded from
of communications between GD&S users. As an orga-
a central data storage server across the network. If the
nization becomes larger or as GIS becomes more dis-
data are not returned to the central server, each person
tributed between offices and branches, the communica-
loses the benefit of having the other's updates. If they
tion system is likely to break down. Long-term storage
both return the data to a central server, one may over-
of local copies of data should be avoided. Upgraded
write the other's changes or there may be two versions
network communications might make it more likely
of the same data on the central server. There may not
that users would work on centralized datasets.
be a way to combine the changes made on each ver-
sion of the data. The changes made in each of the data
Problem: Network protocols
files may not even be known to the other person work-
Description
ing on the same data.
There are a number of network protocols that can
be used over the physical network.
Solutions
Three common scenarios were found for data stor-
Solutions
age and manipulation:
The choice of protocols to lay on top of this phys-
ical layer should be limited to two main options:
Central storage and manipulation
Microsoft networking and TCP/IP. TCP/IP is the
Central storage and local download, manipula-
industry standard for interoperable networking. As
tion, and upload
such it works well with UNIX systems, desktop and
Central storage and local download with no
server PCs, and Macintosh computers, as well as older
replacement of data to the central server.
legacy computers. Microsoft networking has a nar-
rower audience, being confined mostly to Windows
Centralized storage schemes rely on a small num-
PCs. However, many sites reported great success with
ber of network hosts to store the data. Centralized
a product called SAMBA, which allows UNIX-based
storage enables easier implementation of data discov-
servers to communicate with Windows-based PCs
14
Previous Page
