当前位置：网站首页>Azure synapse analytics Performance Optimization Guide (4) -- optimize performance using result set caching

After enabling result set caching , special SQL The pool will automatically cache the query results in the user database , For reuse . such , Subsequent query execution can get results directly from the persistence cache , So there's no need to recalculate . Result set caching improves query performance , And reduce the consumption of computing resources . Besides , Queries that use cached result sets do not occupy any concurrent slots , Therefore, the existing concurrency limit will not be included . For safety reasons , If the data access permission of the accessor user is the same as that of the user who created the cache results , Then the accessor user can only access the cached results . At the database and session levels , By default, the result set cache is “ close ”.

（ Two ） Key orders

1. Enable for user database / Disable result set caching

（1） Check the statistics settings of the database

SELECT name, is_auto_create_stats_on FROM sys.databases

（2） Enable query storage for the database

ALTER DATABASE [database_name]
SET QUERY_STORE = ON;

（3） Enable result set caching for the database

-- Run this command when connecting to the MASTER database

ALTER DATABASE [database_name]
SET RESULT_SET_CACHING ON;

（4） Check the result set cache settings of the database

SELECT name, is_result_set_caching_on
FROM sys.databases;

（5） Enable... For the database Read_Committed_Snapshot Options

Connect to master Run this command while running the database .

ALTER DATABASE MyDatabase
SET READ_COMMITTED_SNAPSHOT ON;

（6） Authority required

A . To set up RESULT_SET_CACHING Options , User needs server level principal login （ Login name created during provisioning ） Or become dbmanager Members of the database role .

B . To set up READ_COMMITTED_SNAPSHOT Options , Users need to have ALTER jurisdiction .

2. Enable for session / Disable result set caching

（1） Example

Using query request_id Inquire about sys.dm_pdw_exec_requests
Medium result_cache_hit Column to see if this query is executed using result cache hits or misses .

SELECT result_cache_hit
FROM sys.dm_pdw_exec_requests
WHERE request_id = 'QID58286'

（2） jurisdiction

Require membership with public roles

3. Check the size of the cached result set

（1） Command statement

DBCC SHOWRESULTCACHESPACEUSED;

remarks
A . Azure Synapse Analytics No server in SQL Pool does not support this syntax .
B .DBCC SHOWRESULTCACHESPACEUSED The command does not use any parameters , And it will return the space used by the database running this command .

（2） jurisdiction

need VIEW SERVER STATE jurisdiction .

（3） Result set

Column	data type	explain
reserved_space	bigint	Total space for database , Unit is KB. This number changes as the cached result set increases .
data_space	bigint	Space for data , Unit is KB.
index_space	bigint	Space for indexing , Unit is KB.
unused_space	bigint	Unused space in reserved space , Unit is KB.

4. Clean cache

from Azure Azure Synapse Analytics Delete all result set cache entries from the database .

（1） grammar

DBCC DROPRESULTSETCACHE
[;]

remarks
This command will empty the result set cache of all queries .
Turning off the result set caching function of the database will also delete all cached results .
Pausing a database with result set caching enabled will not delete cached results .
Azure Synapse Analytics No server in SQL Pool does not support this syntax .

（2） jurisdiction

Require to have DB_OWNER Membership in the fixed server role .

（ 3、 ... and ） What is not cached

After enabling result set caching for the database , The results of all queries will be cached before the cache is full , Except for the following queries ：

Queries with built-in functions or runtime expressions , These expressions are non deterministic expressions , Even if there is no change in the data or query of the base table . for example ,DateTime.Now()、GetDate().
Queries using user-defined functions
Queries using tables with row level security or column level security enabled
The row size in the returned data exceeds 64KB Query for
Return big data （>10GB） Query for

remarks
Some non deterministic functions and runtime expressions can be deterministic for repeated queries of the same data . for example ,ROW_NUMBER().
If the row order in the query result set / Sequences are important to application logic , Please use ORDER BY.
If ORDER BY The data in the column is not unique , Whether the result set cache is enabled or disabled , There's no guarantee that ORDER BY Row order of rows with the same value in the column .

important
The operations of creating a result set cache and retrieving data from the cache are dedicated SQL On the control node of the pool instance . When the result set cache is on , Run to return a large result set （ for example ,>1GB） The query of may cause high bandwidth restriction on the control node , And reduce the overall query response speed on the instance . These queries are usually used in data browsing or ETL Use during operation . To avoid causing pressure on the control node and causing performance problems , Users should turn off the result set cache of the database before running such queries .

The run duration of this query should be the time required to perform the result set cache operation for a query ：

SELECT step_index, operation_type, location_type, status, total_elapsed_time, command
FROM sys.dm_pdw_request_steps
WHERE request_id  = <'request_id'>;

The following is an example output of a query executed with result set caching disabled .

The following is an example output of a query executed with result set caching enabled .

（ Four ） When to use cached results

If all the following requirements are met , The cached result set will be reused for queries ：

The user running the query has access to all tables referenced in the query .
There is an exact match between the new query and the previous query that generated the result set cache .
No data or schema changes have occurred in the table from which the cached result set was generated .

Run the following command to check whether a query has been executed and there are result cache hits or misses . result_cache_hit Column pair cache hit returns 1, Return for cache error 0, The reason why the result set cache is not used , Then return a negative value . For more information , Please check sys.dm_pdw_exec_requests
.

SELECT request_id, command, result_cache_hit FROM sys.dm_pdw_exec_requests
WHERE request_id = <'Your_Query_Request_ID'>

（ 5、 ... and ） Manage cached results

The maximum size of the result set cache for each database is 1 TB. When the basic query data changes , The cached results will be invalidated automatically .

Cache eviction by dedicated SQL The pool is automatically managed according to the following plan ：