I am converting a database from its current character set WE8ISO88591 to AL32UTF8
I am using NLS_LENGTH_SEMANTICS to CHAR so I do not have to increase column lengths.
I follow Oracle instructions:
Except import which takes way too long.
I took a full export of WE8ISO8859p1 Database and now importing it in AL32UTF8.
It is always difficult to prove where the slowness is coming from; but I think it has to do something with NLS_LENGTH_SEMANTICS. On the same
server; if new database was in WE8ISO8859P1, a 5 million row table import took 2 hours; but in Al32UTF8 with NLS_LENGTH_SEMANTICS it is taking 1 day!!!
Any idea how to improve the performance of import.
Character length semantics will slow down the import but the slowdown should be negligible. The slowdown comes from the need to use a globalization function to count characters instead of a much faster comparison of buffer lengths. The character set conversion itself requires much more extra processing time. However, a twelvefold slowdown is very suspicious. You should nail down all tables that take much longer to import and compare their structures to check what they have in common.