On Mon, Jan 25, 2010 at 10:53 PM, nair rajiv
<nair331@gmail.com> wrote:
Hello,
I am working on a project that will take out structured content from wikipedia
and put it in our database. Before putting the data into the database I wrote a script to
find out the number of rows every table would be having after the data is in and I found
there is a table which will approximately have 5 crore entries after data harvesting.
Is it advisable to keep so much data in one table ?
I have read about 'partitioning' a table. An other idea I have is to break the table into
different tables after the no of rows in a table has reached a certain limit say 10 lacs.
For example, dividing a table 'datatable' to 'datatable_a', 'datatable_b' each having 10 lac entries.
I needed advice on whether I should go for partitioning or the approach I have thought of.
We have a HP server with 32GB ram,16 processors. The storage has 24TB diskspace (1TB/HD).
We have put them on RAID-5. It will be great if we could know the parameters that can be changed in the
postgres configuration file so that the database makes maximum utilization of the server we have.
For eg parameters that would increase the speed of inserts and selects.
Thank you in advance
Rajiv Nair
We have several servers that regularly run into records exceeding 50 million records on dual quad core machine with 8 GB RAM and 4 SAS 15K hard disks in RAID 10. If 50 million is the max amount of records that you are looking at, I would suggest not breaking the table. Rather, configure the database settings present in postgresql.conf file to handle such loads.
You already have a powerful machine (I assume it's 16 core, not 16 physical processors), and if configured well, I hope would present no problems in accessing those records. For tuning PostgreSql, you can take a look at pgtune (
http://pgfoundry.org/projects/pgtune/) .
Two changes that I can suggest in your hardware would be to go in for SAS 15K disks instead of SATA if you can do with less capacity, and goign in for RAID 10 instead of RAID 5.
Regards
Amitabh Kant