Hi,
I'm working with at database with more that 500 million records.
Currently I'm using MS SQL server and the performance is awful. If I
were to use BDB, would that improve my query? Is there anybody who has
done some comparative study on BDB vs other Databases with very large
dataset?

regards,
Radzi.

Radzi wrote:
Try IBM's Informix Dynamic Server. For IDS 500 million rows is a small
database (server instance capacity >100PB). You download a free single user
testbed fully functional version from the International Informix Users Group
(IIUG) web site (www.iiug.org) you just have to register for free membership
first. There is user support, similar to this NG, including server tuning
advice available at comp.databases.informix and on several online forums you
can join on the IIUG site. IDS runs on Windoze and most UNIX variants
including Linux and drivers for JDBC, ODBC, ESQL/C, Perl DBD/DBI, and other
popular development environments are available. You can also download the
Software Development Kit free at the IIUG site which includes most of those
interfaces.

Obviously you'll have to buy licenses to rollout multi-user production
databases, but this will get you started and IBM will work with you in many
ways including setting up an eval of the full multi-user releases.

Art S. Kagel

Hi Radzi,

It is difficult to know if Berkeley DB suits your purpose, based on the
information you have given.

Berkeley DB databases can certainly contain substantially more than 500
million records, and maintain performance.

For a good starting point look at the introduction section of the
Reference Guide here:
http://www.oracle.com/technology/doc...b/ref/toc.html

If you have specific questions about the capabilities of Berkeley DB
after reading through the reference guide, we will be glad to help.

Regards,
Alex


Art S. Kagel wrote:

Radzi,

Berkeley DB is based on storing key and value mainly in the server
memory and then sync to hard disk. Because of the fact you can load
entire database into memory it becomes extremely fast database. However
you will not be able to use regular SQL which can make your life more
difficult if you dealing with relational databases. Berkeley DB is
based on storing key and value in mainly the memory of machine and then
syncing to hard disk. Because of the fact you can load entire database
into memory it becomes extremely fast database. However you will not be
able to use regular SQL which can make your life more difficult if you
dealing with relational databases. With 500 million of records you
should definitely try it out, you will have hard time to find a faster
database to handle such a large volume of records.

-Kevin

Radzi wrote: