Disclaimer: I'm in the process of learning SSIS - so forgive me if this is an obvious answer, but I couldn't find anything on searching.
It seems that the Merge transformation is very similar to the Union All transformation, with the following differences:
- Merge can only merge 2 sources of data, whereas Union All supports more (I'm not sure how many - my instructor says 256 but there's no sources to confirm that - and MSDN doesn't mention an upper limit).
- Merge requires inputs to be sorted whereas Union All does not.
Apart from these differences, there doesn't seem to be much else that is different. They are both partially blocking, and from my limited tests, I couldn't see any noticeable difference in performance either.
Everywhere I've searched, I've found that the recommendation is to use "Merge" whenever possible, and to use "Union All" if there are more than 2 data sources, or if the source data is unsorted.
My question is, why? Why would I not just use Union All everywhere? What is the point of Merge?
Once again, I'm sorry if this is a very basic/newbie question, but my instructor doesn't know the answer, and searching hasn't been very fruitful either.
Thanks!
Note: It seems (from some other SO answers that people are mixing up "Merge" with "Merge Join". These are two different things.
Answer
Ok, so the answer is kind of obvious now in hindsight. It seems that the key difference is in the output.
- Merge: Output is Sorted
- Union All: Output is Unsorted
The idea is that if we have sorted input data then why not keep it sorted after our "Merge"/"Union All"? Who knows, we may need sorted data as an input in another transformation down the line, so just use "Merge" if possible so that the data stays sorted.
What's funny is that a whole lot of articles on the internet don't mention this simple difference. Hope this helps another newbie on their way to SSIS mastery!