Well, I guess both are legitimate options, depending on how much flexibility you have with your system design.
The block group approach will definitely result in less load on the system. Having 1 group instead of 1093 is going to be MUCH lighter on the system. It’s also lighter on the database, since you’re only writing the data in one (or a few) transactions, and not in 1000+.
HOWEVER, this doesn’t mean the system can’t handle 1000+ groups, so if you really can’t change the work flow at all, the synchronized execution method should work. The first method, with a block group, also presents some challenges- block groups don’t support polled reads, so you’ll need some way of detecting when all rows are written- assuming that all values change each time, this won’t be too bad, but if some values don’t change it could be a bit difficult.
If using polled reads with the stored procedure groups, each triggered execution could take hundreds of milliseconds (100-500 usually). Times 1093, that’s a lot of work (over a minute). It gets split up a bit, but could still prove difficult to manage.
Ultimately there are lots of considerations. You’ll have to evaluate them and see which one makes more sense. Please feel free to ask any other questions that come to mind, though.
Regards,