Published on August 10, 2017 by Microsoft

This week’s Data Exposed show welcomes Joe Sack into the Channel 9 studio. Joe is a Principal PM on the SQL Server team, and is in the Channel 9 studios today to show us the awesome Adaptive Query Processor technology, the ability of the database to adapt based on customer characteristics in real-time conditions. Joe introduces 3 features of Adaptive QP in SQL Server 2017 and Azure SQL DB.

Today’s AQP topics:

[02:20] – Memory Grant Feedback

[08:50] – Adaptive Joins

[16:20] – Interleaved execution

For more on Adaptive Query Processing, visit our docs page here: 

docs.microsoft.com/en-us/sql/relational-databases/performance/adaptive-query-processing

Leave a Reply

5 Comments on "SQL Server 2017 Adaptive QP"

Notify of
avatar

Joe Sack
Guest
Joe Sack
1 day 17 hours ago

Hi Debdutta,

Adaptive Joins for batch mode helps with issues related to incorrect estimates that impact join algorithm selection. It doesn't change any of the assumptions introduced with the New CE – however if there is a skew related to the New CE that impacts join selection, it may help in that specific case.

In terms of query processing improvements and future strategy – rather than introducing a "new new" CE, we'll be focusing on specific features that address areas that are known to be historically problematic.

Debdutta
Guest
Debdutta
2 days 8 hours ago

In SQL Server 2014 New Cardinality Estimator has been introduced. It makes faster executions of lots of queries, but, we also observe slowness. Base on Microsoft recommendation, we use query hint Trace Flag 9481 (force to use old CE) and it worked as expected.

My question is, does this great Adaptive Join feature solve this problem?

YoshihroKawabata
Guest
YoshihroKawabata
8 days 8 hours ago

Thanks nice presentation. and detail document.Are there any chart of Adaptive Query Plan feature with other DB engines ?Like Oracle, MySQL, SQLite, Azure Cosmos DB, previous SQL Server/Azure SQL Database.For sharing this feature with other DB engine users.   

Joe Sack
Guest
Joe Sack
8 days 16 hours ago

Thanks Carl. Ideally as we add adaptive mechanisms in QP this will also allow customers to reduce the need for explicit hints. We're just getting started in this space, but that is certainly the long-term goal.

CarlD
Guest
CarlD
8 days 17 hours ago

Will be interesting to see how well this works in the coming years. I've found, working with large complex queries that return large numbers of rows that I've always had to heavily hint my queries to get predictable performance.

wpDiscuz