Tag Archives: Performance

SQL Query Store: Compare queries in two periods (execution counts, duration)

If you work with Azure SQL or Microsoft SQL Server 2016+, one of the new features there is the Query Store. It stores useful statistics on queries executed, their execution plans, etc. etc.

One of the scenarios you might find useful is the ability to compare queries executed within two periods:

DECLARE @Period1Start datetime, @Period1End datetime, @Period2Start datetime, @Period2End datetime

SET @Period1Start = '20171212 00:00'
SET @Period1End = '20171213 00:00'
SET @Period2Start = '20171214 00:00'
SET @Period2End = '20171215 00:00'

;WITH period1 AS
(
	SELECT
		query_id,
		AVG(avg_duration) AS avg_duration,
		SUM(count_executions) AS count_executions_total
	FROM sys.query_store_runtime_stats
		INNER JOIN  sys.query_store_runtime_stats_interval ON (query_store_runtime_stats_interval.runtime_stats_interval_id = query_store_runtime_stats.runtime_stats_interval_id)
		INNER JOIN sys.query_store_plan ON query_store_plan.plan_id = query_store_runtime_stats.plan_id
	WHERE
		(sys.query_store_runtime_stats_interval.start_time >= @Period1Start)
		AND (sys.query_store_runtime_stats_interval.end_time <= @Period1End)
	GROUP BY query_id
),
period2 AS
(
	SELECT
		query_id,
		AVG(avg_duration) AS avg_duration,
		SUM(count_executions) AS count_executions_total
	FROM sys.query_store_runtime_stats
		INNER JOIN  sys.query_store_runtime_stats_interval ON (query_store_runtime_stats_interval.runtime_stats_interval_id = query_store_runtime_stats.runtime_stats_interval_id)
		INNER JOIN sys.query_store_plan ON query_store_plan.plan_id = query_store_runtime_stats.plan_id
	WHERE
		(sys.query_store_runtime_stats_interval.start_time >= @Period2Start)
		AND (sys.query_store_runtime_stats_interval.end_time <= @Period2End)
	GROUP BY query_id
)
SELECT
		query_store_query.query_id,
		period1.avg_duration AS period1_avg_duration,
		period2.avg_duration AS period2_avg_duration,
		CASE WHEN period1.count_executions_total IS NOT NULL THEN (period2.avg_duration - period1.avg_duration) * 100.0 / period1.avg_duration ELSE NULL END AS avg_duration_increase_percent,
		period1.count_executions_total AS period1_count_executions_total,
		period2.count_executions_total AS period2_count_executions_total,
		CASE WHEN period1.count_executions_total IS NOT NULL THEN (period2.count_executions_total - period1.count_executions_total) * 100.0 / period1.count_executions_total ELSE NULL END AS count_execution_increase_percent,
		query_sql_text
	FROM period2
		LEFT JOIN period1 ON period1.query_id = period2.query_id
		LEFT JOIN sys.query_store_query ON query_store_query.query_id = period2.query_id
		LEFT JOIN sys.query_store_query_text ON query_store_query_text.query_text_id = query_store_query.query_text_id
	--ORDER BY period2.count_executions_total DESC
	ORDER BY count_execution_increase_percent DESC
	--ORDER BY avg_duration_increase_percent DESC

References

SQL DMV: Most Expensive Queries, Missing Indexes

Database Management Views and their usage in SQL performance tuning scenarios are well known. There are plenty posts with many different versions of the Most Expensive Queries query. As I don’t want to look for the right one again and again I archive my favorite one here:

-- Most expensive queries
SELECT TOP 20
	SUBSTRING(qt.TEXT, (qs.statement_start_offset/2)+1,
		((CASE qs.statement_end_offset
			WHEN -1 THEN DATALENGTH(qt.TEXT)
			ELSE qs.statement_end_offset
			END - qs.statement_start_offset)/2)+1)
		AS query_text,
    db.name AS [db_name],
    qs.total_elapsed_time/1000 AS total_elapsed_time_ms,
    qs.total_elapsed_time/qs.execution_count/1000 AS average_elapsed_time_ms,
    qs.last_elapsed_time/1000 AS last_elapsed_time_ms,
    qs.execution_count,
    qs.total_worker_time/1000 AS total_worker_time_ms,
    qs.total_worker_time/qs.execution_count/1000 AS average_worker_time_ms,
    qs.last_worker_time/1000 AS last_worker_time_ms,
    qs.last_execution_time,
    qs.total_logical_reads,
	qs.last_logical_reads,
    qs.total_logical_writes,
	qs.last_logical_writes
    --qp.query_plan
    FROM sys.dm_exec_query_stats qs
        CROSS APPLY sys.dm_exec_sql_text(qs.sql_handle) qt
        CROSS APPLY sys.dm_exec_query_plan(qs.plan_handle) qp
        LEFT JOIN sys.databases db ON (qt.dbid = db.database_id)
    -- WHERE db.name='DatabaseName'
    -- ORDER BY qs.total_logical_reads DESC
    -- ORDER BY qs.total_logical_writes DESC
    -- ORDER BY qs.last_worker_time DESC -- CPU time (active)
    -- ORDER BY qs.last_elapsed_time DESC -- clock time (including locks, etc.)
    -- ORDER BY qs.total_worker_time DESC
	ORDER BY qs.total_elapsed_time DESC

…and for the same purpose the one for missing indexes (Please don’t add these indexes without further analysis of the overall impact!):

SELECT
	mid.statement
	  ,migs.avg_total_user_cost * (migs.avg_user_impact / 100.0) * (migs.user_seeks + migs.user_scans) AS improvement_measure,OBJECT_NAME(mid.Object_id),
	  'CREATE INDEX [missing_index_' + CONVERT (VARCHAR, mig.index_group_handle) + '_' + CONVERT (VARCHAR, mid.index_handle)
	  + '_' + LEFT (PARSENAME(mid.statement, 1), 32) + ']'
	  + ' ON ' + mid.statement
	  + ' (' + ISNULL (mid.equality_columns,'')
		+ CASE WHEN mid.equality_columns IS NOT NULL AND mid.inequality_columns IS NOT NULL THEN ',' ELSE '' END
		+ ISNULL (mid.inequality_columns, '')
	  + ')'
	  + ISNULL (' INCLUDE (' + mid.included_columns + ')', '') AS create_index_statement,
	  migs.*, mid.database_id, mid.[object_id]
	FROM sys.dm_db_missing_index_groups mig
		INNER JOIN sys.dm_db_missing_index_group_stats migs ON migs.group_handle = mig.index_group_handle
		INNER JOIN sys.dm_db_missing_index_details mid ON mig.index_handle = mid.index_handle
	WHERE migs.avg_total_user_cost * (migs.avg_user_impact / 100.0) * (migs.user_seeks + migs.user_scans) > 10
	ORDER BY migs.avg_total_user_cost * migs.avg_user_impact * (migs.user_seeks + migs.user_scans) DESC