Loading data into TimesTen

Question

Hi,
We are looking into options that we have to load data into TimesTen. The data would be in data files.
We are looking at three options -- Custom jdbc program, Custom PL/SQL procedure or ttBulkCp.
Which would be a better option in terms of performance ? If we consider a data size of 100 million records, which one would perform better ?
Does anybody have any performance benchmarks for any of the same ?
Thanks and Regards,
Pritom

Answer 1

ttBulkCp is usually pretty fast (especially in direct mode). However, it is a generc program so it does have to parse each input line. Depending on the im0plementation and file format, custom code might have to do less parsing and so might be a little faster. However, Java/JDBC is significantly slower thsan C/ODBC so YMMV. Custom C/ODBC code would likely be somewhat faster than ttBulkCp but whether the difference is worth the effort is hard to say. Custom PL/SQL is unlikely to be faster than direct mode ttBulkCp but I have never compared them.
I don't personally know of any benchmark numbers that compares these three methods (or any subset of them).
Note that as long as the loading mechanism is reasonably efficient then the hardware spec and TimesTen configuration will also be very important in obtaining optimal load performance.
Chris