August 29, 2020 at 1:46 am
So I had a performance issue with a developer attempting to import over a billion records into a hash table (no indexes no keys etc...) using similar code "INSERT INTO TABLENAME SELECT COLUMN A, COLUMN B FROM QUERY"
It took about 5.5 hours total, but with the below hack I was able to drop it down to 3.5 hours.
By including the "INSERT INTO TABLENAME WITH (TABLOCK) SELECT COLUMN A, COLUMN B FROM QUERY" it appears starting with SQL Server 2016+, inserts will be using parallelism compared to the old serial method of inserting records. The maxdop I had for the server is 8.
Tung Dang
Azure and SQL Server DBA Contractor / Consultant
DataZip
August 29, 2020 at 6:48 am
What is your question?
The absence of evidence is not evidence of absence.
Martin Rees
You can lead a horse to water, but a pencil must be lead.
Stan Laurel
Viewing 3 posts - 1 through 2 (of 2 total)
You must be logged in to reply to this topic. Login to reply