Dear All,

Please advice me how can I change the default 100 records to 20000 when bdbreconfig6.1 loads data into Oracle tables?

Any Oracle level information also highly appreciated.

Thanks in advance.

Eric.

I may be corrected on this but I think it loads one record at a time but only updates the display every hundred.

trying using this bw option

-- -set RDS_FULL=value to define max. nr. rows transfered between client and server as one block (good for imports/exports)

Indeed the rds_full setting will increase performance for this however, I would not recommend setting this to 20000 (see this thread RDS_FULL (http://www.baanboard.com/baanboard/showthread.php?s=&threadid=108) ) and keep in mind that the display will still count the records by increments of 100.

Yes, be careful not to put RDS_FULL at too high a level. It will stop Baan from issuing a database commit too often - thus improving performance - but it may lead to memory problems.

Hi,

There is also a bw option MAX_INPUT_ROWS (I think that is correct, but am away from my Baan notes, also see the MAX_FETCH_ROWS option) this will inform Baan to load the set amount of records in each read.

But again do not put it too high, I have used it to 5000 and its been ok, but it can cause problems.

Hope this helps


Steve

Dear All,

I have tried 6000 and it worked then I changed to 20000 then also it worked. After that we had some changes on database side(oracle) like larger temporary space for sys,root,bsp etc. Larger roll back segment etc..

Porting set is 6.1c.6.06 for HP8000_9000 and DB(oracle 8.1.7.2)

Now it would not load more than 100 records at a time.

This was the db_resource then,

ora_max_array_insert:20000
ora_max_array_fetch:20000
oracle_client_home:/opt5/oracle8i
ora_column_format:7

Now we have stopped the ARCH in Oracle to see whether it would improve or not. We stopped the db,rc.stop and even did server shutdown to clear any resource issues. no results.

When we run this migration of c3 to c4 no one else is using the server so we are not bothered about OLTP or other issues.
We would like to finish the 68 million record in 5 hours than 22 hours!.

Thanks for all replies and any new information that give us an idea.

Eric.

Good. :)

Glad to know it can go higher will add that to my notes when I get back. Please let me know if you have problems before the end, and how long it takes.


Steve

Steve

Can you post more information on MAX_INPUT_ROWS ?

thanks,

Eric

Hi,

Sorry, its the ora_max_array_insert that you used, its been a while since I used it and I don't have my notes with me.

You could try Baan Support, there are a few notes on it.


Steve.

Dear All,

I have tested combination and found it reads file in 20000 but the write is still 800 records per second that worries me.

any ideas let me know.

thanks,Eric

Dear All,

By using the following db_resource
dbsinit:021
ora_init:0111000
ora_max_array_insert:20000
ora_max_array_fetch:20000
ssts_set_rows:20000
rds_full:20000
ora_timeout:{999,999,999,999,999}
oracle_client_home:/opt5/oracle8i
ora_column_format:7
nls_sort:binary
epc_disable:true
lock_retry:0

I could get the speed of 20000 records per second but write went up to 800 from 100 records. Could some one help me to get a higher value in writing ?

thanks, Eric.

Hi Eric Thomas,
I don't know which table you are trying to reconfig. But we recently migrated from IVc to IVc4. we have 28 million records in tfgld418 and baan has added a field and the field is also part of index. So, this is what we did. our backend is informix. since it is tfgld418 table all the records are static. So, we went live on dec 1998. So from 1998 onto last month we did unload the data from the production server using informix unload tool. we did 3 months at a time parallel during day time. I created a company 555 in 4c4. we created a table tfgld418555 in informix with new schema in informix. created with good first extent and next extent sizes. then i just changed the data files from bdbpre6.1 by using awk script to include 0 in the correct position. the data file before running awk script is as follows :
701|1|2|03/26/1999|55771|1|XPR|99000120|800|700021|3||2|5|63|701001002199903260557710000100000
After running the awk script I removed "/" from date and added 0 after 1 and before XPR.
now the file looks as :
701|1|2|03/26/1999|55771|1|0|XPR|99000120|800|700021|3||2|5|63|701001002199903260557710000100000

I have installed baan4c4 in different BSE and baan4c in different BSE. so i changed my BSE to baan4c4 BSE and ran bdbpost6.1 using the above file as my input file into company 555. This process took 2 days. No harm done.. Users are working as a charm in production companies. On sunday, we came and logged off users and did last unload from last month and later. did a bdbpost6.1 on 555. Renamed tfgld418555 to tfgld418800 in informix.
Migration of big table complete without bringing the system down for 22 hours. no need of changing parameters... worked like a charm.

Hi Csekhar,

We have done a trail on live copy to convert c3 to c4 and in the trail run tfgld419 failed and the number of records was 68 million and we could get back into table in 5 hours. then we changed oracle parameters,etc to make it even faster and also in the process we would have changed the db_resource also.
Then it turned to 100 records per second. After booking a case with baan they modified the change package combination to run faster and it is 800 records on insert. The approach you took is good but we have several companies running on baan it would not be feasible to do the manual work of the size we have.

We are planning X'mas weekend to convert the companies so it is possible to do on baan gui but the speed would save holidays..

thanks,Eric.

Eric,

Some other to things to consider besides tweaking the db_resource file could be your client-server setup. If you are running in a multi-tier environment you could opt to install a Baan workstation environment - for the migration usage only - or indeed if you are running in a cluster setup, you could make sure both database and application are running on the one node. With this in mind you can configure a BEQ (bequeth) connection rather than using the TCP Oracle listener which will speed up things as well. And finally, don't use the Baan GUI for exporting/importing data but rather the $BSE/bin command line utilities.