Ben Domenico wrote:
Hi Shishir,
I'm forwarding your specific questions to the THREDDS technical staff,
but, in the meantime, I'll point you to our status pages where you
should be able to find answers to some of your questions and the
software used to implement some parts of the system.
<http://www.unidata.ucar.edu/projects/THREDDS/>
and the technical status page:
<http://www.unidata.ucar.edu/projects/THREDDS/tech/index.html>
Please let me know if you have further questions.
--Ben
******************************************************
Ben Domenico Unidata Deputy Director
http://www.unidata.ucar.edu P.O. Box 3000
ben@xxxxxxxxxxxxxxxx Boulder, CO 80307
(303)497-8631 FAX: (303)497-8690
--On Wednesday, June 18, 2003 1:51 PM -0700 "Shishir S. Bharathi"
<shishir@xxxxxxx> wrote:
Hi,
I'm working at USC-Information Sciences Institute for the Globus
group. I'm looking at some aspects of metadata catalogs and was
interested in obtaining more information about the usage and
architecture of metadata catalogs in the THREDDS project. I was going
through this paper in the Journal of Digital Information and yours was
the contact information provided. Could you answer some of the
questions
that I had or point me to the right person to ask ?
1. How is the data stored in the PICats ? Are you using relational or
XML based databases ? Is your schema a general one or particular to the
PICat in question ?
Picats = catalogs are simply lists of online datasets. they have
metadata, but the data itself is not stored in the catalogs. So we dont
use databases at this level at all.
The Catalog schema is a DTD that applies to all catalogs.
2. What is the query mechanism to query a PICat ? What kind of
queries
can be performed ?
We are working on query services that are built on top of catalogs. The
main kinds of queries we will support are geographic and time bounding
boxes, as well as keyword seraching.
What is the exact mapping stored in a PICat ? Can I
perform an Attribute list/Condition list --> Logical file mapping ?
A catalog stores the URL of the actual dataset. Not sure what you mean
by mapping.
3. What is the mapping at the level of the PICat servers ?
Not sure what you mean by mapping.
4. The document says that PICat servers monitors a list of PICats. Is
the information from the PICats pulled or do the PICats update the PICat
server ? Is this update a soft state update (meaning the update expires
after a certain time) ?
It is pulled. Catalogs are passive XML documents.
We expect to add "expire" semantics to catalogs in the next (0.7) version.
5. I'm a bit confused by the statement that a PICat server can be
distributed over many hosts ? How is this done ?
I'm not sure what that sentence means either. A catalog can point to
datasets on many servers, is what i guess.
6. The PICats servers propagate changes to the other servers ?
What is
the information propagated ?
We have not yet implemented this, but it will probably exchange enough
information to make distributed searching as efficient as possible. I
dont know if we will do a replication or not.
7. The document also mentions that by connecting to any PICat server
the client gets (approximately) the same information. What does this
mean
? Is is only the update latency or is there some information that is not
propagated to all the servers ?
The idea is that these search servers will be connected, so that the
query will effectively be sent to all the servers. SO you dont have to
contact more than one server to get an answer to your query.
There will of course be latency issues. However, it will be a loosely
coupled system that will not make many guarentees.
Thanks and regards,
Shishir
--
