[UW-GIS-L] [Questions about requirements for ArcGIS_Server 9.2
Harvey Greenberg
hgreen at u.washington.edu
Thu May 24 13:15:55 PDT 2007
We intend to upgrade our data server, and we have many questions about
hardware and software purchasing, and about setup.
What we have: We are running ArcIMS 9.0 on coal.ess.washington.edu
(WinXP, IIS, Tomcat, Java) and ArcSDE 9.0 running under
WindowsServer2003 with MS SQLServer on a 172 address
(crusty.ess.washington.edu).
We want to install the new ArcGIS Server 9.2 on a machine on the 128
domain. (We won't risk touching our old installations of ArcIMS and
ArcSDE until we have duplicated their functionality.) We aspire to
serve maps and to serve (like SDE) large datasets in read-only mode to
ArcGIS Desktop users. We are prepared to use passwords and other means
to limit access to the capacity of a small machine. Does ArcGIS Server
really replace SDE? It seems to more or less replace IMS, but the
SDE-like part is still a mystery:
What we have tried: We have installed the .NET version of ArcGIS Server
and (free) SQL Express on an XP machine for testing (using IIS for web
services). The map server (functionality more or less replacing IMS) was
fairly easy to set up and use (though it's mysteriously broken at the
moment) but we can't get the data server ("GIS resources" -- which would
presumably replace SDE functionality) to work. Also, the ArcGIS Server
installation never seemed to indicate a necessity for a RDBMS so the SQL
Express is not in use.
If I want to serve a dataset ("GIS resource") via ArcGIS Server like I
do currently with SDE, does it really have to be in a DB or at least in
a file-GeoDB, or can it just be, say, a lone TIF lying there in a directory?
Do we still need a RDBMS like SQLserver or Oracle in order to serve
large data sets? I read somewhere that ESRI file geodatabases are
faster than SDE/SQL. Is this even true? If so, true across scales?
If we don't need a DBMS, should we try RedHat linux? That would involve
using the java version of ArcGIS Server rather than .NET. We hope that
would not be as "complicated" (flaky) as running Java and Tomcat with
IMS 9.0. We have a general fear of a more difficult installation so
wonder if anyone has experience or advice. If we use Windows, do we want
Server2003 or can we get away with XP and IIS? (Can we please not use
Vista?)
Do we need MS Server2003 to run SQLserver? (I think we do....
In any case, perhaps a better question for techsupport than UW GIS-L?)
How much money do we need for that? (Is there a UW price for
SQL? We inherited the SDE machine with Server2003 and SQL so we don't
know what that costs. Again, perhaps a better question for techsupport?)
What are the chokepoints on our system? Would it be CPU speed, number
of cores, bandwidth, local disk access (or fileserver access if we
choose that route?), memory speed, memory size, or some subtle feature
of our chipset? How do people monitor the chokepoints on a functioning
GIS Server system?
We have been shopping the Dell website, and found a sweet core2Duo
Optiplex for cheap. The Precision line seems beefier than we need, or
is there something about Xeons or their chipset that is important for
this sort of server?
Thank you for reading this entire message,
harvey and charles
======================================================================
hgreen at u.washington.edu 374 Johnson Hall 206-685-7981
http://gis.ess.washington.edu/
Charles Kiblinger http://staff.washington.edu/cek/
cek at u.washington.edu v206.685.7981 f206.543.0489
Johnson 374 | Box 351310 | University of Washington | Seattle WA 98195
More information about the Uw-gis-l
mailing list