On 4/13/06, Jun Li <[EMAIL PROTECTED]> wrote: > Reading Oracle model is very slow, should I define the connection parameter? > Will it help?
I cannot really help you there. Perhaps there are parameters that affect this, but I don't know them . > I used the batch mode for inserting data, but for a large amount of data > (about 80,000) in one table, it doesn't help much. It took over 12hrs to get > about 6,000 records inserted on SQL Server. However, DataToDatabaseSink > works well for other small tables. I tried to set batch size from 100 to > 1000, but it doesn't make much difference in my case. Is there anyway to > improve this? Speed of batch mode largely depends on the batch size (which depends on the amount of memory you have and the type of data, e.g. usually you can specify 10000 easily). And even more important is the order in which you insert. Say, if you have three tables A, B, C, then an XML where you have <A ...> <B ...> <C ...> <A ...> <B ...> <C ...> ... will effectively turn batch mode off because in batch mode you can only insert into a single table. Hence, a change of tables will result in the commit of the batch. So you should order your XML such that <A ...> ... <A ...> <B ...> ... <B ...> <C ...> ... <C ...> and this if possible in foreign key order with the referenced tables before the referencing tables. > BTW, you said "create table in Oracle", I thought this function hasn't been > finished, was my memory serve me wrong or you've finished this stuff. I think you confuse that with something else, e.g. CREATE DATABASE ? Tom
