SQL SERVER – Audit Script to Get CPU and Memory Information with MAXDOP Guidelines

If you are reading this blog regularly the name of Dominic Wirth may not be new for you. He has in the past sent some very interesting scripts for audit for our blog and I am including his new script in this blog which is about how to Get CPU and Memory Information with MAXDOP Guidelines.

 

Here are few of the previous blog post by Dominic Wirth. He is truly SQL genius.

Let us see today’s script which talks about some very interesting information about PCU and Memory and along with that it also provides guidelines for MAXDOP.


/*
====================================================================
Author: Dominic Wirth
Date created: 2019-04-18
Date last change: -
Script-Version: 1.0
Tested with: SQL 11.0.6020 and above

Description: This script shows important information regarding
Processor and Memory configuration of the server
as well as some configuration recommendations
for query parallelism. These recommendations are a
summary of MANY recommendations found on the
internet. The valid setting for a server ALWAYS
depends on its usage so there is no right/wrong.
====================================================================
*/
DECLARE
@sqlMajor INT
,@sqlMinor INT
,@cpuLogicals INT
,@cpuCores INT
,@numaConfig NVARCHAR(60)
,@allocType NVARCHAR(120)
,@stmt NVARCHAR(MAX)
,@paramDef NVARCHAR(500);
DECLARE @serverMemory TABLE (SQL_MinMemory_MB BIGINT, SQL_UsedMemory_MB BIGINT, SQL_MaxMemory_MB BIGINT, Server_Physical_MB BIGINT, SQL_AllocationType NVARCHAR(60));
DECLARE @cpuInfo TABLE (VirtualMachineType NVARCHAR(60), NUMA_Config NVARCHAR(60), NUMA_Nodes INT, Physical_CPUs INT
,CPU_Cores INT, Logical_CPUs INT, Logical_CPUs_per_NUMA INT, CPU_AffinityType VARCHAR(60)
,ParallelCostThreshold_Current INT, MAXDOP_Current INT, MAXDOP_Optimal_Value NVARCHAR(60), MAXDOP_Optimal_Reason NVARCHAR(1024));
-- ========================================================================================================================================================================
-- Load SQL information
SELECT @sqlMajor = CAST((@@MicrosoftVersion / 0x01000000) AS INT), @sqlMinor = CAST((@@MicrosoftVersion / 0x010000 & 0xFF) AS INT);

-- ========================================================================================================================================================================
-- Load CPU configuration
--
IF @sqlMajor > 12
BEGIN
SELECT @stmt = 'SELECT @numaConfig = softnuma_configuration_desc, @allocType = sql_memory_model_desc FROM sys.dm_os_sys_info;'
,@paramDef = '@numaConfig NVARCHAR(60) OUTPUT, @allocType NVARCHAR(120) OUTPUT';
EXEC sp_executesql @stmt, @paramDef, @numaConfig = @numaConfig OUTPUT, @allocType = @allocType OUTPUT;
END
ELSE
SELECT @numaConfig = 'UNKNOWN', @allocType = 'UNKNOWN';
--
INSERT INTO @cpuInfo (VirtualMachineType, NUMA_Config, NUMA_Nodes, Logical_CPUs, CPU_AffinityType, MAXDOP_Current, ParallelCostThreshold_Current)
VALUES (
(SELECT virtual_machine_type_desc AS VirtualMachineType FROM sys.dm_os_sys_info)
,@numaConfig
,(SELECT COUNT(memory_node_id) FROM sys.dm_os_nodes WHERE memory_node_id < 64)
,(SELECT COUNT(scheduler_id) FROM sys.dm_os_schedulers WHERE scheduler_id < 255)
,(SELECT affinity_type_desc AS AffinityType FROM sys.dm_os_sys_info)
,CAST((SELECT [value] FROM sys.configurations WHERE [name] = 'max degree of parallelism') AS INT)
,CAST((SELECT [value] FROM sys.configurations WHERE [name] = 'cost threshold for parallelism') AS INT)
);
--
-- Get CPU Information
IF OBJECT_ID('tempdb..#cpu_output') IS NOT NULL
DROP TABLE #cpu_output;
CREATE TABLE #cpu_output ([output] VARCHAR(255));

INSERT INTO #cpu_output ([output]) EXEC xp_cmdshell 'wmic cpu get DeviceId,NumberOfCores,NumberOfLogicalProcessors /format:csv';
-- Example [output]: CLVDODWI01,CPU0,4,8
--
-- Remove empty lines, header line and CrLf
DELETE FROM #cpu_output WHERE REPLACE(REPLACE(RTRIM(ISNULL([output],'')), CHAR(10), ''), CHAR(13), '') = '';
DELETE FROM #cpu_output WHERE [output] LIKE '%NumberOfCores,NumberOfLogicalProcessors%';
UPDATE #cpu_output SET [output] = REPLACE(REPLACE([output], CHAR(10), ''), CHAR(13), '');
--
--SELECT * FROM #cpu_output;
--
-- Load information of CPU cores and logical processors
SELECT @cpuCores = 0, @cpuLogicals = 0;
SELECT @cpuCores += PARSENAME(REPLACE([output], ',', '.'), 2)
,@cpuLogicals += PARSENAME(REPLACE([output], ',', '.'), 1)
FROM #cpu_output;
--
UPDATE @cpuInfo SET Physical_CPUs = (SELECT COUNT(*) FROM #cpu_output), CPU_Cores = @cpuCores, Logical_CPUs = @cpuLogicals;
--
IF OBJECT_ID('tempdb..#cpu_output') IS NOT NULL
DROP TABLE #cpu_output;
--
UPDATE @cpuInfo SET Logical_CPUs_per_NUMA = Logical_CPUs / NUMA_Nodes;
--
-- Calculate best MAXDOP setting according guidelines of a Microsoft Premier Filed Engineer
-- > Hyper-threading enabled: Should not be 0 and should not be greater than half the number of logical processors
-- > Processor Affinity set: Should not be more than the number of cores available to the SQL Server instance
-- > NUMA: Should be no more than the number of cores per NUMA node to avoid expensive foreign memory access that occurs,
-- when a task needs to use memory that does not belong to its NUMA node.
-- > Generic: If you are unsure of the above values then a generic setting should not be more than 8.
-- So if you have more than 8 logical processors you should set this value to a maximum of 8.
UPDATE @cpuInfo
SET
MAXDOP_Optimal_Value = '0 - ' + CAST((CASE WHEN Logical_CPUs > 8 THEN 8 ELSE Logical_CPUs END) AS NVARCHAR)
,MAXDOP_Optimal_Reason = 'GENERIC: Not more than the amount of logical CPUs and not more than 8';
--
UPDATE @cpuInfo
SET
MAXDOP_Optimal_Value = '0 - assigned cores to the SQL Server instance'
,MAXDOP_Optimal_Reason = 'CPU AFFINITY: Not more than the amount of assigned CPUs to the SQL Server instance'
WHERE CPU_AffinityType = 'MANUAL';
--
UPDATE @cpuInfo
SET
MAXDOP_Optimal_Value = '1 - ' + CAST(Logical_CPUs_per_NUMA AS NVARCHAR)
,MAXDOP_Optimal_Reason = 'NUMA NODES: Not 0 and not more than the number of cores per NUMA node'
WHERE NUMA_Nodes > 1;
--
UPDATE @cpuInfo
SET
MAXDOP_Optimal_Value = '1 - ' + CAST((Logical_CPUs / 2) AS NVARCHAR)
,MAXDOP_Optimal_Reason = 'HYPER-THREADING: Not 0 and not be greater than half the number of logical processors'
WHERE Logical_CPUs / 2 = CPU_Cores;
--
-- ========================================================================================================================================================================
-- Load Memory configuration

INSERT INTO @serverMemory (SQL_MinMemory_MB, SQL_UsedMemory_MB, SQL_MaxMemory_MB, Server_Physical_MB, SQL_AllocationType)
VALUES (
CAST((SELECT [value] FROM sys.configurations WHERE [name] = 'min server memory (MB)') AS BIGINT)
,(SELECT cntr_value / 1024 FROM sys.dm_os_performance_counters WHERE counter_name = 'Total Server Memory (KB)')
,CAST((SELECT [value] FROM sys.configurations WHERE [name] = 'max server memory (MB)') AS BIGINT)
,(SELECT (total_physical_memory_kb / 1024) AS total_physical_memory_mb FROM sys.dm_os_sys_memory)
,@allocType
);
--
-- ========================================================================================================================================================================
-- Show informations
--
SELECT VirtualMachineType, NUMA_Config, NUMA_Nodes, Physical_CPUs, CPU_Cores, Logical_CPUs, Logical_CPUs_per_NUMA, CPU_AffinityType FROM @cpuInfo;
SELECT ParallelCostThreshold_Current, MAXDOP_Current, MAXDOP_Optimal_Value, MAXDOP_Optimal_Reason FROM @cpuInfo;
SELECT * FROM @serverMemory;

When I ran above script I got the following output on my test machine.

While the script is amazing and I love every bit of it. I must tell you one thing, whenever you follow any guideline or any suggestion, always please make sure that you have taken proper approval from your manager. If you are not sure what you are doing, just do not do it. I have seen quite a few people changing some of the important settings of their database without knowing what actually do. So yea, if there is no one around to talk to you, please send me an email or leave a comment.

Reference: Pinal Dave (https://blog.sqlauthority.com)

MAXDOP, Parallel, SQL CPU, SQL DMV, SQL Memory, SQL Scripts, SQL Server
Previous Post
SQL SERVER – Difference Between Azure Data Studio and SQL Server Management Studio
Next Post
SQL SERVER – Docker Volume and Persistent Storage

Related Posts

2 Comments. Leave new

  • Anandan Kanagarajan
    December 22, 2019 9:52 pm

    Very useful Audit query and a must have in DBA’s tool box. Thanks for both Dominic & Paul…

    Reply
  • Very useful. I assigned 15 GB as Max Memory to SQL server. After restart, SQL server slowly reaches to 15 GB and doesn’t release it until next restart. How can I check how much actually SQL server using. Is there any way that I can find SQL server utilizing certain amount than 15GB

    Reply

Leave a Reply

Menu
Exit mobile version