August 16, 2019 at 12:23 am
Trying to narrow the problem down, but a rough outline of the overall process is:
August 16, 2019 at 12:38 am
just curious how the updates are being handled?
August 16, 2019 at 12:53 am
Lookup to determine new / existing, conditional split for changed / unchanged, and and OLE DB Command for the actual update, like all the other data flows in the ETL. (which aren't slow)
August 16, 2019 at 12:58 am
Don't use OLEDB Command for updates ... unless your number of updates is tiny. It's dog slow.
Instead, direct all of your update rows to (yet) another staging table (as INSERTs), and at the end of your dataflow, issue a single UPDATE from staging to target.
The absence of evidence is not evidence of absence
- Martin Rees
The absence of consumable DDL, sample data and desired results is, however, evidence of the absence of my response
- Phil Parkin
August 16, 2019 at 1:38 am
Well, this is a big one (for this data set), at ~200-450K rows. So, maybe not _tiny_, but certainly not huge, compared to a lot of others out there.
August 16, 2019 at 2:44 pm
Well, this is a big one (for this data set), at ~200-450K rows. So, maybe not _tiny_, but certainly not huge, compared to a lot of others out there.
The OLEDB Command does its updates one row at a time. 500K rows will take a significantly longer time to process in this manner than in the way I described, I'd put money on it.
The absence of evidence is not evidence of absence
- Martin Rees
The absence of consumable DDL, sample data and desired results is, however, evidence of the absence of my response
- Phil Parkin
Viewing 6 posts - 16 through 20 (of 20 total)
You must be logged in to reply to this topic. Login to reply