Hi All,

I am running Create Table from Seq.Dump and its running very very slow. I am uploading table ticpr200 which has approx.53000000 and in 12 hours only 12000000 records are uploaded. I put RDS_FULL=60000 in my bw client but no use.

Urgent help is required.

Thanks in advance

Hi,

Forget to mention. I am working on Baan-IV c4, HP-Unix 11

What database?

If it's oracle you could be extending your Datafiles or writing archive logs like crazy.

If it's Informix, you may have similar problems, but also informix has DD locking issues and if you're trying to do anything simultaneously you may be locking the inf DD tables

Also setting RDS_FULL that high could cause you more problems than good.

I never go higher than 100, i've gotten through 15 million rows of a 16 million row table before just to have it error out with a corrupt record (which was caused by the driver not liking my high RDS_FULL value.

it should also match up with your array_fetch parameters.

so for oracle
RDS_FULL=100 SSTS_SET_ROWS=100 ORA_MAX_ARRAY_INSERT=100


Oh - did you chose "create rows before index"? because that's a big one as well


Dave

Thanks Dave,

I am on oracle 9.2.0. I have not chosan `Create Row before index'.

Can you please explain me what is the meaning of SSTS_SET_ROWS=100 ORA_MAX_ARRAY_INSERT=100.

Should I extned the datafile.

regards

For mass import you should use 'create row before index'. This will accelerate your import.

I am on oracle 9.2.0. I have not chosan `Create Row before index'.


Choosing that will improve your performance significantly.


Can you please explain me what is the meaning of SSTS_SET_ROWS=100 ORA_MAX_ARRAY_INSERT=100.


Those are all documented in the Driver manual - if you don't have one, i suggest you get one!!


Should I extned the datafile.



depends on how much freespace you have in the tablespace...
your datafile should not be set to autoextend though, size the tablespaces accordingly ahead of time.


Dave

Hi

Thanks Dave but my problem is system performance. I tried bdbpost also but no improvement.

Thanks in advance

Manoj

If you don't see the relationship between what i said above and system performance.. then you probably have more problems on your system than I can help with over forums.

Dave

We have also some serious performance problems within importing data. Using the session "create table from sequential dump" it takes over 48 hours to import a data dump with a size of approx 43 GB (Informix 10, Solaris 9, quite fast machine with 16 GB of RAM and 6 Sparc-CPUs). We have split the import process in different single processes and afaik 'Create Row Before Index' had been also activated... Especially the tables with more than 14 million rows need very long to import.

Does there any other option exist to accelerate the import process? What if we would import the data by using Informix tools and not the Baan tools (bdbpost, create table from sequential dump)?

peturba

We have split the import process in different single processes


Check your informix locking while importing, Especially in informix 10. Informix wants to lock it's internal system tables for each separate import so you may be hurting yourself more than helping.

Create Sequential Dump of table just calls bdbpost, the only thing you gain by going to bdbpost directly is the ability to put it in cron or use a schell script if you'd like.

You may want to benchmark using a separator vs not using a separator as well.

Dave

Hello,

Turn off the AUDIT-option on the tables.
That also will increase the performance.

G.N.

Thanks All,

My problem solved. As my database is Oracle so I upload the data using SQL Loader and then I built index.

Dear Friends,

Baan IVc4 has its own limitations , however hard you try it will give the same performance. Even a system which has capability of handling 64 threads at a time with 32GB memory will not improve the performance.

Instead of bdbpost please try Informix dbexport/dbimport method ( HPL , High performance Loader of informix ).

Regards
Shiv Sharma

bdbpost would be good option to load 53 million records , give paramter "r 100000" and switch off "Audit" on Database.