We have crated a really small configuration from scratch (just 2 catalogs). These 2 catalogs are populated from text files by searching matches and updating them if no match was found.
I have noticed (with the performance profiler) significant write speed difference with 2 catalogs. Writing to the second catalog took 97,8% of all the time of the program. 2nd catalog have almost the same number of attributes (around 10) and all of them are of simple types, like string, date or number.
What could affect writing performance by that much considering that:
1. catalogs does not have complex type attributes
2. they are not hierarchical
3. there are no links from or to the catalogs
4. there are more than 200K+ records in each of them
5. infobase was put on a RAM-drive to eliminate any disk subsystem problems
6. Infobase was relatively small (200M in size)
Could variable length strings affect performance?
Is there any maximum attributes summary length?
If there is more than one index could it degrade write performance by much?
Something else, that I have forgotten to mention?
Could it be platform version related problem? I was testing on 8.3.12.1586
Please, help me with any ideas. It is so painfully slow now, that is going to take some days to populate Infobase with data.
Also I have noticed that while testing typically 1c app consumes more and more memory at the begging of ejection, then reaches around 450-600Mb of RAM and starts decreasing use of memory, staying at around 90Mb of RAM.
The 1C:Enterprise developers forum