Adaptive Join Behavior

Grant Fritchey explains adaptive joins in SQL Server 2017:

Currently the adaptive join only works with columnstore indexes, but according to Microsoft, at some point, they will also work with rowstore. The concept is simple. For larger data sets, frequently (but not always, let’s not try to cover every possible caveat, it depends, right), a hash join is much faster than a loops join. For smaller data sets, frequently, a loops join is faster. Wouldn’t it be nice if we could change the join type, on the fly, so that the most effective join was used depending on the data in the query. Ta-da, enter the adaptive join.

First, the statistics are used at compile time for the tables we’re joining. Based on those statistics, a row target it set. Below that threshold, a loops join will be used. Above that threshold, a hash join. The way the row count is determined is that the operator will always build the hash table. With the hash table built and loaded, it will know how many rows it has. If it’s going to do a loops join, the hash table is tossed and a loops join commences. If the threshold has been passed on the row counts and it’s going to do a hash join, it already has the hash table built and proceeds to do a hash join. It’s easy to understand. Let’s see it in action.

Click through to see it in action.  It’s not a world-changer yet, but as it becomes available to rowstore queries (without the filtered, empty columnstore index trick), I think people will come to appreciate it.

Related Posts

Batch Mode Normalization

Paul White digs into batch mode normalization and its consequences for performance: I mentioned in the introduction that not all eight-byte data types can fit in 64 bits. This fact is important because many columnstore and batch mode performance optimizations only work with data 64 bits in size. Aggregate pushdown is one of those things. There are […]

Read More

Comparing CAST and CONVERT Performance

Max Vernon runs a performance test of CAST versus CONVERT: This post is a follow-up to my prior post inspecting the performance of PARSE vs CAST & CONVERT, where we see that PARSE is an order of magnitude slower than CONVERT. In this post, we’ll check if there is a similar difference between using CAST or CONVERT. But just to be clear, CONVERT offers […]

Read More

Categories

February 2018
MTWTFSS
« Jan Mar »
 1234
567891011
12131415161718
19202122232425
262728