How do databases process queries in parallel (more than one query at a time) while still keeping the correct order (queries are performed in the same order that they are submitted)? Is there a kind of queue structure? thanks.
asked Aug 20, 2015 at 7:41
-
f the question were about concurrent execution, rather than parallel execution, problems relating to concurrency control and reproducibility have been studied since the 1970s. Look up "ACID Transactions", "concurrency control", and "serializability" for articles concerning this topic.– Walter MittyCommented Aug 20, 2015 at 11:37
Add a comment
|
1 Answer
Answers relevant to SQL Server.
If talking about query priority:
For SQL Server there is by default a FIFO principle adhered to. The first query to run gets priority.
However with the resource governor it is possible to not only limit the resources assigned to your sql server instances, it is also possible to create workload groups which can prioritize certain connections to be able to run before others.
If talking about parallelism:
For SQL Server it does this by running parallel streams, that are fed information by a parallel page supplier, and in their turn have their information fed through a gather streams operator into a stream aggregate operator, which gathers the information provided by the separate streams to then be passed on further.
In other words, for SQL Server this is done by gathering the processed data at the end, and bundling it a new. If needed they can then be sorted or filtered afterwards.
See this simple talk article for more details.
