Hello
I am researching how to store the data for easy 'user-driven'
reporting (where I do not need to develop application for
every user request).
The data will typically be number ranges and text strings with
unique Id for each row
I hope there will be a lot of data :-).
So in that anticipation I am looking for a way
to allow
SQL/ODBC access to the data
but in a way that each table resides on more than one
PG server
for example:
table 1 lives in 3 PG instances (I can partition the data by date range)
table 2 lives in the same 3 instances plus another one (because it's
bigger)
and I would like users to be able to issue SQL from within ODBC that
joins them.
I do not mind if I have to for example name tables like
tb1_srv1 and so on
but some how then have the joined with the global view -- if such
functionality exists.
That is, the users with SQL/ODBC will only be doing reads -- not updates
and some how they have to see a 'single' tables that resides on multiple
hosts.
I do not mind the complexity for the inserts (because I can program for
it) -- but selects need to be easy from things like MS Access and
Crystal reports, or for more sophisticated users from packages like 'R'
The reason why I think the data will not fit into one database,
is because I just do not have money for servers (everything is coming
out
of my small pocket) so I just want to deploy inexpensive computers
but add them as I get more data to serve.
I looked at Hypertable but there is no ODBC driver for it.
Alternatively I also looked making the host operating systems into
a single Image OS. Which appears to be possible with
http://www.kerrighed.org/wiki/index.php/Main_Page
However, I do not know if PG will even run there.
Obviously top-of the line performance for me is not necessary
but has to be not unusable.
thank you in advance
--
Vlad P
author of C++ ORM http://github.com/vladp/CppOrm/tree/master
--
http://www.fastmail.fm - A fast, anti-spam email service.