Hi Guys,

Has any body ever split a table into multiple parts to improve performance ? How is it done and whether it is effective.

The Integration table has crossed 50,000,000 records, there are other such tables.

Checking whether we can avoid archiving ?

Would also be interested in knowing if this sort of thing can be done with SQL Server as the back end??

Thnx
sk

Has any body ever split a table into multiple parts to improve performance ? How is it done and whether it is effective.
We have considered Oracle Partitioning and discussed but have not implemented it yet. More info @ Partitioning in Oracle (http://www.oracle.com/technology/products/bi/db/11g/pdf/partitioning-11g-whitepaper.pdf)
The Integration table has crossed 50,000,000 records, there are other such tables.

Checking whether we can avoid archiving ?
Do you really need to save all those records? if not its better to purge them.

Is it Baan Application dependent or Database dependent ??
:confused:

This is what we are doing in Baan Vc in our inf_storage6.2 file. This might work for you.
tfgld106:180:T:group:0200:10:INITIAL 1800000 NEXT 180000 LOCK row [round robin in tfgld10601dbs, tfgld10602dbs, tfgld10603dbs]
tfgld106:180:I1:group:0200:10:DBSPACE tfxxxxxx04idx
tfgld106:180:I2:group:0200:10:DBSPACE tfxxxxxx05idx
tfgld106:180:I3:group:0200:10:DBSPACE tfxxxxxx06idx
tfgld106:180:I4:group:0200:10:DBSPACE tfxxxxxx07idx
tfgld106:180:I5:group:0200:10:DBSPACE tfxxxxxx08idx
tfgld106:180:I6:group:0200:10:DBSPACE tfxxxxxx01idx
tfgld106:180:I7:group:0200:10:DBSPACE tfxxxxxx02idx
tfgld106:180:I8:group:0200:10:DBSPACE tfxxxxxx03idx

Hi Nicholas,

Were you using it the normal way and have later splitted the tables ? and has this improved the performance ??

Yes we were the "normal way", as in the tfgld106 table was only residing in a single dbspace. What happens with Round robin, is it writes record number 1 in the 1st dbspace, record number 2 in the 2nd dbspace and 3 record in the 3rd dbspace and so forth.

We try to keep a single dbspace & associated chunks on a single LUN, to assist with tracking where our performance hits are. We were originally 100% on IO. After we split this we dropped down to 33% on IO.

Hope that helps.

Hi Nicholas,

How did u measure the IO activity for a specific table ? Could u tell me the command ?

I have the following queries :
1) how many records do you have in tfgld106.
2) Is it your largest table.
3) How many extents do you have for each of the three dbspace for tfgld106
4) For howmany tables have u done that ?