[RFC 004] Operation Review Comment 12

I'm going to go through all the questions and answer as I can. The stats


from our motherlode server are in the answers to questions 7 and 8.


1) Unidata has been serving much of our real-time IDD data with the
OPeNDAP C++ netCDF server for the last four years or so. We have also
developed our own server (the OPeNDAP Aggregation Server) and client
library (netCDF-Java) that are used by a number of sites inside and
outside the Unidata community. The Unidata IDV data analysis and
visualization package supports OPeNDAP data access and is also used
widely.


2) We use OPeNDAP in data visualization apps and support OPeNDAP in our
netCDF-Java library. We do not use OPeNDAP in all our applications or
for all the types of data we use. This is partly because OPeNDAP servers


are not available for all the data types we use and that some of our
applications have other methods of accessing remote data.


3) We work with a variety of data access systems besides OPeNDAP
including ADDE and OGC WCS.


4) We have not found OPeNDAP servers to be difficult to install or
maintain. Our support load has not increased much because of our use of
OPeNDAP servers.


5) Most of our OPeNDAP use has been very satisfactory in terms of
performance. As the size of the requested dataset increases, of course,
the longer the time to access the datasets. The ability to request a
subset of a dataset greatly alleviates this problem. We have not used
the OPeNDAP-G implementation but that is another way around this issue
for those accessing/providing very large datasets.


6) I don't believe OPeNDAP entails much, if any, greater challenge than
making data available in other ways. Once you have a machine that can
handle access to your data collection (disk, network, etc) via FTP or
HTTP, adding OPeNDAP access is not much harder.


7 & 8) We do not have information on the volume of data handled but here


is what we have for the number of users per month and the number of data


sets accessed per month. (You can probably ignore the last row. The
difference between it and the top row is how unique data requests is
counted. The bottom one counts each unique subset request seperately.)
Sorry we haven't done 2005 data yet. This is for 2004:


Jan Feb Mar Apr May Jun Jul Aug
Sep Oct Nov Dec
----- ---- ---- ---- ---- ---- ---- ---- ----
---- ---- ---- ----
# of Unique 3554 5787 8611 4180 4028 8833 3809 2036
1995 1324 1146 1138
datasets
(no CE)


# of Unique 439 483 539 464 372 436 323 414
448 454 443 338
Users


# of Unique 33079 39542 48324 35616 35688 186252 31222 32278
35677 39014 36356 31741 datasets (w/ CE)



Hope this is helpful,