Transcript StarCITE Engineering Update

MarketView Database Tuning Best Practice
StarCite Engineering
May 13, 2009
CONFIDENTIAL
21-Jul-15
1
Agenda
• Identify the hot-spot
– Long Running DML
– Inefficient Indexes
– More dig into (tempdb, disk usage, etc)
• Improve the Long Running DML
• Improve the Inefficient indexes
• TempDB improvement
• Production Data
• Some hot topics
• Further topics
CONFIDENTIAL
21-Jul-15
2
Identify the hot-spot - Long Running DML
• Powerful DMV & DMF
Dynamic Management View & Dynamic Management Function
CONFIDENTIAL
21-Jul-15
3
Long Running DML
Identify top 10 query by CPU cost
CONFIDENTIAL
21-Jul-15
4
Long Running DML
Identify top 10 query by elapsed time
CONFIDENTIAL
21-Jul-15
5
Long Running DML
MKV tools can identify to specific table
SELECT TOP 50
qs.total_worker_time/qs.execution_count as [Avg CPU Time],
qt.text as query_text,
qt.dbid,
qt.objectid
FROM sys.dm_exec_query_stats qs
cross apply sys.dm_exec_sql_text(qs.sql_handle) as qt
Where qt.text like '%table_name%'
ORDER BY [Avg CPU Time] DESC
CONFIDENTIAL
21-Jul-15
6
Inefficient indexes
Identify the missing Index
CONFIDENTIAL
21-Jul-15
7
Inefficient indexes
Identify the most costly unuage index
CONFIDENTIAL
21-Jul-15
8
Dig into
• TempDB
• Data file fragmentation
• Log file fragmentation
• Disk usage
CONFIDENTIAL
21-Jul-15
9
TempDB
Provide it a best environment!
Temp Store
Procedure
Temp Table
Put it on faster disk(raid5).
• TempDB
Table variables
Row version
Provide
separated disk for tempdb, and avoid
the working files sit with os files.
Internal Table
When you have multiple CPUsCursor
may increase
tempdb file numbers can help.
Increase tempdb initial size it depends?
CONFIDENTIAL
21-Jul-15
10
TempDB (Space Used By)
User objects
These are explicitly created by user sessions and are tracked in system catalog. They
include the following:
 Table and index.
 Global temporary table (##t1) and index.
 Local temporary table (#t1) and index.
 Session scoped.
 Stored procedure scoped in which it was created.
 Table variable (@t1).
 Session scoped.
 Stored procedure scoped in which it was created.
Internal objects
These are statement scoped objects that are created and destroyed by SQL Server to
process queries. These are not tracked in the system catalog. They include the
following:
 Work file (hash join)
 Sort run
 Work table (cursor, spool and temporary large object data type (LOB) storage)
As an optimization, when a work table is dropped, one IAM page and an extent is saved
to be used with a new work table.
There are two exceptions; the temporary LOB storage is batch scoped and cursor
worktable is session scoped.
Version Store
This is used for storing row versions. MARS, online index, triggers and snapshot-based
isolation levels are based on row versioning. This is new in SQL Server 2005.
Free Space
This represents the disk space that is available in tempdb.
CONFIDENTIAL
21-Jul-15
11
TempDB (Space Used By)
User objects
+
Total Space Usage
Internal objects
+
Version Store
+
Select
SUM (user_object_reserved_page_count)*8 as user_objects_kb,
SUM (internal_object_reserved_page_count)*8 as internal_objects_kb,
SUM (version_store_reserved_page_count)*8 as version_store_kb,
SUM (unallocated_extent_page_count)*8 as freespace_kb
From sys.dm_db_file_space_usage
Where database_id = 2
Here is one sample output (with space
user_objets_kb internal_objects_kb
---------------- -------------------8736
128
in KBs).
version_store_kb
-----------------64
freespace_kb
-----------448
Free Space
CONFIDENTIAL
21-Jul-15
12
TempDB
Some of the common issues with tempdb are as follows:
•
Running out of storage space in tempdb.
•
Queries that run slowly due to the I/O bottleneck in tempdb.
•
Excessive DDL operations leading to a bottleneck in the system tables.
•
Allocation contention.
CONFIDENTIAL
21-Jul-15
13
TempDB
look for multiple counters to cross check the validity of your findings.
•
PhysicalDisk Object: Avg. Disk Queue Length represents the average number of physical read and write
requests that were queued on the selected physical disk during the sampling period. If your I/O system is overloaded,
more read/write operations will be waiting. If your disk queue length frequently exceeds a value of 2 during peak usage
of SQL Server, then you might have an I/O bottleneck.
•
Avg. Disk Sec/Read is the average time, in seconds, of a read of data from the disk. Any number
Less than 10 ms - very good
Between 10 - 20 ms - okay
Between 20 - 50 ms - slow, needs attention
Greater than 50 ms – Serious I/O bottleneck
•
Avg. Disk Sec/Write is the average time, in seconds, of a write of data to the disk. Please refer to the guideline in
the previous bullet.
•
Physical Disk: %Disk Time is the percentage of elapsed time that the selected disk drive was busy servicing read
or write requests. A general guideline is that if this value is greater than 50 percent, it represents an I/O bottleneck.
•
Avg. Disk Reads/Sec is the rate of read operations on the disk. You need to make sure that this number is less than
85 percent of the disk capacity. The disk access time increases exponentially beyond 85 percent capacity.
•
Avg. Disk Writes/Sec is the rate of write operations on the disk. Make sure that this number is less than 85 percent
of the disk capacity. The disk access time increases exponentially beyond 85 percent capacity.
CONFIDENTIAL
21-Jul-15
14
TempDB
When you have identified an I/O bottleneck, you can address it by
doing one or more of the following:
•
Check the memory configuration of SQL Server. If SQL Server has been configured with
insufficient memory, it will incur more I/O overhead.
•
Increase I/O bandwidth.
* Add more physical drives to the current disk arrays and/or replace your current disks
with faster drives. This helps to boost both read and write access times. But don't add
more drives to the array than your I/O controller can support.
* Add faster or additional I/O controllers. Consider adding more cache (if possible) to your
current controllers.
•
Examine execution plans and see which plans lead to more I/O being consume. It is
possible that a better plan (for example, index) can minimize I/O. If there are missing
indexes, you may want to run Database Engine Tuning Advisor to find missing indexes
CONFIDENTIAL
21-Jul-15
15
TempDB
1. The following DMV query can be used to find which batches/requests are
generating the most I/O.
select top 5 (total_logical_reads/execution_count) as avg_logical_reads,
(total_logical_writes/execution_count) as avg_logical_writes, (total_physical_reads/execution_count) as
avg_phys_reads, Execution_count, statement_start_offset as stmt_start_offset, sql_handle,
plan_handle from sys.dm_exec_query_stats order by (total_logical_reads + total_logical_writes) Desc
2. You can also identify I/O bottlenecks by examining the latch waits. These
latch waits account for the physical I/O waits when a page is accessed
for reading or writing and the page is not available in the buffer pool.
Select wait_type, waiting_tasks_count, wait_time_ms from sys.dm_os_wait_stats where
'PAGEIOLATCH%' order by wait_type
CONFIDENTIAL
wait_type like
21-Jul-15
16
TempDB
1. Put it on faster disk(raid5).
2. Provide seperated disk for tempdb
3. Increase tempdb file number. A suggestion number is 1 processer 1 file
4. Increase tempdb initial size
CONFIDENTIAL
21-Jul-15
17
Disk Term
•
Hard Drive: HDD or disk, non-volatile storage device which stores
digitally encoded data on rapidly rotating platters with magnetic
surfaces.
•
Platter: Two or more per HDD.
•
Cylinder: All tracks which can be accessed by the heads while the
access arms are stationary.
•
Track: Concentric rings on a platter.
•
Sector
– 1. Wedge-shaped sections of a platter, classically 64 sectors per track.
– 2. Bits which lie at the intersection of a track & a sector, usually 512 bytes.
(This is a standard definition & one we will use most.)
•
File Allocation Unit: Some integral number of 512 byte sectors which
are treated as a unit by the OS.
CONFIDENTIAL
21-Jul-15
18
Disk Issue Implicit
•
Output of diskpar (Windows 2000 Resource Kit) (Above Windows 2000 should use
diskpart.exe)
C:\>diskpar -i 0
---- Drive 0 Geometry Infomation ---Cylinders = 12161
TracksPerCylinder = 255
SectorsPerTrack = 63
BytesPerSector = 512
DiskSize = 100027630080 (Bytes) = 95393 (MB)
---- Drive Partition 0 Infomation ---StatringOffset = 32256
PartitionLength = 49319424
HiddenSectors = 63
PartitionNumber = 1
PartitionType = de
–
By default, for years Windows instantiated 63 hidden sectors in all new partitions.
–
These hidden sectors contain the master boot record (MBR).
–
Note the typos:
•
“StatringOffset” instead of “StartingOffset”.
•
“Infomation” instead of “Information”
CONFIDENTIAL
21-Jul-15
19
Disk Issue Implicit
CONFIDENTIAL
21-Jul-15
20
Disk Issue Implicit (File Allocation Unit)
• fsutil fsinfo ntfsinfo c:
• Output for default NTFS format
C:\>fsutil fsinfo ntfsinfo c:
NTFS Volume Serial Number :
Version :
0x3a16ff9d16ff5879
3.1
Number Sectors :
0x000000000a2397ff
Total Clusters :
0x00000000014472ff
Free Clusters :
0x000000000025b76a
Total Reserved :
0x00000000000051f0
Bytes Per Sector :
512
Bytes Per Cluster :
Bytes Per FileRecord Segment
4096
: 1024
Clusters Per FileRecord Segment : 0
Mft Valid Data Length :
0x0000000007c90000
...
CONFIDENTIAL
21-Jul-15
21
Disk Issue Implicit (File Allocation Unit)
Extent
Data Page
64KB
8KB
……
So we need change the file allocation unit to 64KB that could reduce I/O significantly.
CONFIDENTIAL
21-Jul-15
22
Disk file fragmentation
File A in
DataBase A
File A in
DataBase A
•
Two types of disk Reading :
1.
sequential reading (Index seek)
2.
Random reading(Index scan)
File B in
DataBase B
•
Tuning:
1.
Increase initial data file size
2.
Set correct data file increasing size
3.
Schedulable clean up file fragmentation
Unused space
File B in
DataBase B
CONFIDENTIAL
21-Jul-15
23
Log file fragmentation
• More log file cause data fragmentation
• Small size log file.
• Less log file (VLF). A suggestion number is 5
• Schedule backup log file.
CONFIDENTIAL
21-Jul-15
24
Improve the Long Running DML
CONFIDENTIAL
21-Jul-15
25
Improve the Long Running DML
Select dbuser1_.id as id190_,
…
…
From
DB_SITE_USERS dbsiteuser0_,
DB_USERS dbuser1_,
SITE_MODULES sitemodule2_,
DB_WORKSHOPS dbworkgrou3_,
DB_USER_ROLES dbuserrole4_,
DB_ROLE_MODULES dbrolemodu5_,
DB_ROLES dbrole6_,
WORKGROUP_RESOURCES workgroupr7_,
RESOURCES resource8_,
RESOURCE_ITEMS resourceit9_
Where
dbsiteuser0_.SITE_ID=35
AND dbsiteuser0_.DELETED_FLAG=0
AND dbuser1_.USER_ID=dbsiteuser0_.USER_ID
AND dbuser1_.DELETED_FLAG=0
AND sitemodule2_.SITE_ID=35
AND sitemodule2_.MODULE_ID=17
AND sitemodule2_.DELETED_FLAG=0
AND dbworkgrou3_.SITE_ID=35
AND dbworkgrou3_.DELETED_FLAG=0
AND dbrole6_.SITE_ID=35
AND dbrole6_.ROLE_ID=dbrolemodu5_.ROLE_ID
AND (
dbrole6_.NAME in (
'RFP EscalatiON','BrAND Manager','Executive Management'
)
)
AND dbrole6_.DELETED_FLAG=0
AND dbuserrole4_.WORKGROUP_ID=dbworkgrou3_.WORKGROUP_ID
AND dbuserrole4_.SITE_USER_ID=dbsiteuser0_.SITE_USER_ID
AND dbuserrole4_.DELETED_FLAG=0
AND dbrolemodu5_.ROLE_ID=dbuserrole4_.ROLE_ID
AND dbrolemodu5_.ACTIONS % 8-4>=0
AND dbrolemodu5_.DELETED_FLAG=0
AND workgroupr7_.WORKGROUP_ID=dbuserrole4_.WORKGROUP_ID
AND workgroupr7_.DELETED_FLAG=0
AND resourceit9_.RESOURCE_ID=workgroupr7_.RESOURCE_ID
AND resourceit9_.DELETED_FLAG=0
AND resource8_.id=resourceit9_.RESOURCE_ID
AND resource8_.DELETED_FLAG=0
AND ((resourceit9_.RESOURCE_ITEM_XREF_ID=1381
or resourceit9_.ALL_RESOURCE_ITEM_XREFS=1
) AND resourceit9_.RESOURCE_ITEM_TYPE_ID=1
or ( resourceit9_.RESOURCE_ITEM_XREF_ID in (90)
or resourceit9_.ALL_RESOURCE_ITEM_XREFS=1)
AND resourceit9_.RESOURCE_ITEM_TYPE_ID=2
)
CONFIDENTIAL
21-Jul-15
26
Improve the Long Running DML
•
SELECT u.user_name as userName, u.email as email
•
FROM RESOURCE_ITEMS as ri with(nolock)
•
JOIN WORKGROUP_RESOURCES as wr with(nolock)
•
ON ri.deleted_flag=0 AND wr.deleted_flag = 0
•
AND ri.resource_item_type_id IN( 1,2)
•
AND ri.resource_item_xref_id IN(1381, 90)
•
AND ri.site_id=35
•
AND ri.resource_id = wr.resource_id
•
JOIN DB_USER_ROLES ur with (nolock)
•
ON ur.workgroup_id = wr.workgroup_id AND ur.deleted_flag = 0
•
JOIN DB_ROLES r with (nolock)
•
ON r.role_id = ur.role_id
•
JOIN DB_ROLE_MODULES rm with (nolock)
•
ON r.deleted_flag = 0 AND rm.deleted_flag =0 AND r.role_id = rm.role_id
•
AND r.role_id in ('35\18','35\81','35\82') AND r.site_id = 35
•
AND (rm.MODULE_ID=17 or all_modules = 1) AND (rm.ACTIONS % 8-4>=0)
•
JOIN DB_SITE_USERS su ON su.site_user_id = ur.site_user_id AND su.deleted_flag = 0
•
JOIN DB_USERS u ON u.user_id = su.user_id AND u.deleted_flag = 0
CONFIDENTIAL
21-Jul-15
27
Improve the Long Running DML
How about the
index?
CONFIDENTIAL
21-Jul-15
28
Improve the Long Running DML
CONFIDENTIAL
21-Jul-15
29
Improve the Long Running DML
•
How can we improve the long running reading query?
1.
Restrict the size of input. The leftmost table selection is most important factor for the
performance of query.
2.
Reduce the size of output. Only read you need read data from the output tables.
3.
Try you best not involve the String, Date comparison or like. If you must, please make
sure correct use and avoid the index scan happen.
4.
Read the query plan generated by sql server. For small or medium size query try you best
use “INNER LOOP” or “HASH MATCH”. For large size query try you best use “MERGE
JOIN”.
5.
If customer of your system can accept some level data inconsistency, please put
with(nolock) table hint that will reduce read lock contention. If customer of your system
cannot accept data inconsistency and your sql server version is 2005, you also can try
SNAPSHOT ISOLATION level (VERSION DATA) but that will come without free (each
query will occupy more memories than normal other isolation level).
6.
Join type need be carefully considered. Avoid “OUTTER JOIN” and “RIGHT JOIN”.
7.
Performance killer “Table Spooler”, “Stream Aggregation”. Count(*) will impose the “Table
Spooler” and “Stream Aggregation”, try you best use other way for implementation.
Holy Grail Algorithm
8.
MSDN will be your best assistance.
CONFIDENTIAL
21-Jul-15
30
Improve the Long Running DML
•
How can we improve the long running create, update or delete query?
1. Restrict the size of data you want to change in the table. Try you best
control data changes within the row level, otherwise sql server may
escalate the lock level from row level to table level or schema level.
With(rowlock) table hint is not 100 percent true.
2. Large data change need to be batch mode (within one transaction).
Multiple transactions data changes will impose contention and even
further dead lock.
3. Change only your need change data. Because that possibly will not
reflect changes to some NONE-CLUSTER indexes.
4. Avoid concurrent long running changing transactions. If really need that,
separate them into small transactions (NESTED TRANSACTION).
CONFIDENTIAL
21-Jul-15
31
Improve the Long Running DML
MKV performance bottleneck due to the long running transactions.
Application
transaction
enlist
Db transaction
enlist
Sql server 2005
enlist
Db transaction
Db transaction
Update table A
Table A
1
Update table B
Update table C
*
Table B
1
*
Table C
FINE-GRAINED TRANSACTION or COARSE-GARINED TRANSACTION
?
CONFIDENTIAL
21-Jul-15
32
Improve the Inefficient indexes
• Identify the slowness reason from indexes perspective
• Identify the slowness reason from contention perspective
• Reduce the contention of update by unused indexes
• Apply the trace configuration for diagnostic purpose if deadlock found.
• Group review & Performance testing.
CONFIDENTIAL
21-Jul-15
33
Dynamic Management Views and Functions
Dynamic management views and functions return server state information that
can be used to monitor the health of a server instance, diagnose problems,
and tune performance.
• Index Related Dynamic Management Views and Functions
– sys.dm_db_index_operational_stats
– sys.dm_db_index_usage_stats
– sys.dm_db_missing_index_details
– sys.dm_db_missing_index_groups
– sys.dm_db_index_physical_stats
– sys.dm_db_missing_index_columns
– sys.dm_db_missing_index_group_stats
CONFIDENTIAL
21-Jul-15
34
About the Missing Indexes Feature Indexes perspective
The missing indexes feature uses dynamic management objects and
Showplan to provide information about missing indexes that could enhance
SQL Server query performance.
• Dynamic Management Objects
– sys.dm_db_missing_index_group_stats
– sys.dm_db_missing_index_groups
– sys.dm_db_missing_index_details
– sys.dm_db_missing_index_columns
Returns information
summaryinformation
detailed
about a
information
specific
about
the
database
a missing
group
about
table
ofindex;
missing
missing
columns
for
indexare
indexes,
example,
that
groups,
missing
such
it returns
for
asan
example,
the
the
index.
group
name
the performance
identifier
and
identifier
and of
thethe
identifiers
table of
improvements
all
where
missing
the index
indexes
that
is missing,
could
that are
be
gained
contained
and
theby
columns
in
implementing
that group.
and column
a
specific
types
that
group
should
of missing
make up
indexes.
the
missing index.
CONFIDENTIAL
21-Jul-15
35
Examples - Indexes perspective
• Find the 10 missing indexes with the highest anticipated improvement for user
queries
select top 10 d.*, s.avg_total_user_cost, s.avg_user_impact, s.last_user_seek,
s.unique_compiles from sys.dm_db_missing_index_group_stats s,
sys.dm_db_missing_index_groups g,
sys.dm_db_missing_index_details d
where s.group_handle = g.index_group_handle
and d.index_handle = g.index_handle
order by s.avg_total_user_cost * s.avg_user_impact * (s.user_seeks + s.user_scans) desc
CONFIDENTIAL
21-Jul-15
36
Examples - Indexes perspective
• Suggested index columns & usages
SELECT mig.*, statement AS table_name,
column_id, column_name, column_usage
FROM sys.dm_db_missing_index_details AS mid
CROSS APPLY sys.dm_db_missing_index_columns (mid.index_handle)
INNER JOIN sys.dm_db_missing_index_groups AS mig ON mig.index_handle =
mid.index_handle
ORDER BY mig.index_group_handle, mig.index_handle, column_id;
CONFIDENTIAL
21-Jul-15
37
Indexes perspective - Improve
• The result set of sys.dm_db_missing_index_details returns this
information in the equality_columns, inequality_columns, and
included_columns columns.
• The result set returned by sys.dm_db_missing_index_columns
returns this information in its column_usage column.
• Create Index by Using Missing Index Information
– List the equality columns first (leftmost in the column list).
– List the inequality columns after the equality columns (to the right of
equality columns listed).
– List the include columns in the INCLUDE clause of the CREATE INDEX
statement.
– To determine an effective order for the equality columns, order them
based on their selectivity; that is, list the most selective columns first.
CONFIDENTIAL
21-Jul-15
38
Limitations of the Missing Indexes
Feature
• It is not intended to fine tune an indexing configuration.
• It cannot gather statistics for more than 500 missing index groups.
• It does not specify an order for columns to be used in an index.
• For queries involving only inequality predicates, it returns less accurate cost
information.
• It reports only include columns for some queries, so index key columns must be
manually selected.
• It returns only raw information about columns on which indexes might be missing.
• It does not suggest filtered indexes.
• It can return different costs for the same missing index group that appears multiple
times in XML Showplans.
• It does not consider trivial query plans.
CONFIDENTIAL
21-Jul-15
39
Index Perspective – Balance (1)
declare @dbid int
select @dbid = db_id()
select 'object' = object_name(object_id),index_id
,'user reads' = user_seeks + user_scans + user_lookups
,'system reads' = system_seeks + system_scans + system_lookups
,'user writes' = user_updates
,'system writes' = system_updates
from sys.dm_db_index_usage_stats
where objectproperty(object_id,'IsUserTable') = 1
and database_id = @dbid order by 'user reads' desc
CONFIDENTIAL
21-Jul-15
40
Index Perspective – Balance (2)
declare @dbid int
select 'object'=object_name(o.object_id), o.index_id
, 'usage_reads'=user_seeks + user_scans + user_lookups
, 'operational_reads'=range_scan_count + singleton_lookup_count
, range_scan_count, singleton_lookup_count, 'usage writes' = user_updates
, 'operational_leaf_writes'=leaf_insert_count+leaf_update_count+ leaf_delete_count
, leaf_insert_count,leaf_update_count,leaf_delete_count
, 'operational_leaf_page_splits' = leaf_allocation_count
, 'operational_nonleaf_writes'=nonleaf_insert_count + nonleaf_update_count + nonleaf_delete_count
, 'operational_nonleaf_page_splits' = nonleaf_allocation_count
from sys.dm_db_index_operational_stats (@dbid,NULL,NULL,NULL) o ,sys.dm_db_index_usage_stats u
where objectproperty(o.object_id,'IsUserTable') = 1
and u.object_id = o.object_id and u.index_id = o.index_id
order by operational_reads desc, operational_leaf_writes, operational_nonleaf_writes
CONFIDENTIAL
21-Jul-15
41
Index Perspective – Statistics
• Slow down the performance: INSERTs, UPDATEs, and DELETEs
• Whether or not the indexes will be used, decided by Query Optimizer.
– Selectivity: the percentage of rows in a table that are returned by a query
• Index statistics – Used by Query Optimizer for selectivity info
– Density (Conceptually reverse to selectivity)
– DBCC SHOW_STATISTICS (table_name, index_name)
CONFIDENTIAL
21-Jul-15
42
Contention perspective - List Indexes with
the Most Contention
• sys.dm_db_index_operational_stats
• sys.indexes
USE G3ProdMKV_MarketView
declare @dbid int
select top 10 dbid=database_id,
objectname=object_name(s.object_id),indexname=i.name, i.index_id,partition_number,
row_lock_count,row_lock_wait_count,row_lock_wait_in_ms,
[block %]=cast (100.0 * row_lock_wait_count / (1 + row_lock_count) as numeric(15,2)),
[avg row lock waits in ms]=cast (1.0 * row_lock_wait_in_ms / (1 + row_lock_wait_count) as
numeric(15,2))
from sys.dm_db_index_operational_stats (@dbid, NULL, NULL, NULL) s, sys.indexes i
where objectproperty(s.object_id,'IsUserTable') = 1 and i.object_id = s.object_id
and i.index_id = s.index_id order by row_lock_wait_count desc
CONFIDENTIAL
21-Jul-15
43
Contention perspective - Highest Average
CPU Time and Highest Execution Counts
The following sample scripts lists the top 10 statements by average
CPU time.
– sys.dm_exec_query_stats
– sys.dm_exec_sql_text(sql_handle)
SELECT TOP 10
qs.total_worker_time/qs.execution_count as [Avg CPU Time],qs.execution_count,
SUBSTRING(qt.text,qs.statement_start_offset/2,
(case when qs.statement_end_offset = -1
then len(convert(nvarchar(max), qt.text)) * 2
else qs.statement_end_offset end - qs.statement_start_offset)/2)
as query_text, qt.dbid, dbname=db_name(qt.dbid), qt.objectid
FROM sys.dm_exec_query_stats qs
cross apply sys.dm_exec_sql_text(qs.sql_handle) as qt
ORDER BY [Avg CPU Time] DESC
CONFIDENTIAL
21-Jul-15
44
Contention perspective - Contention of
update by unused indexes
• sys.indexes
• sys.objects
• sys.dm_db_index_usage_stats
select top 10 object_name(i.object_id), i.name
from sys.indexes i, sys.objects o
where i.index_id NOT IN (select s.index_id
from sys.dm_db_index_usage_stats s
where s.object_id=i.object_id and i.index_id=s.index_id and database_id = db_id())
and o.type = 'U' and o.object_id = i.object_id
order by object_name(i.object_id) asc
CONFIDENTIAL
21-Jul-15
45
Contention perspective – Deadlock
diagnostic
• Enables the specified trace flags.
• Syntax
– DBCC TRACEON ( trace# [ ,...n ][ , -1 ] ) [ WITH NO_INFOMSGS ]
trace# - Is the number of the trace flag to turn on.
n - Is a placeholder that indicates multiple trace flags can be specified.
-1 - Switches on the specified trace flags globally.
WITH NO_INFOMSGS - Suppresses all informational messages.
• Examples
– DBCC TRACEON (1222);
– DBCC TRACEON (1222, -1);
– DBCC TRACEON (1222, 3206, -1);
CONFIDENTIAL
21-Jul-15
46
Impact by indexes
• Whenever you add one index into the table. Should view the impact to
the INSERT, UPDATE and DELETE by the index you added .
Set statistics time on
Set statistics io on
Updated sql before/after
Set statistics time off
Set statistics io off
If updated sql slow than before, you should know the table read/write
ratio. If read most, it should be acceptable. But if write most, it
definitely impose bad impact overall.
CONFIDENTIAL
21-Jul-15
47
TempDB improvement
• Tempdb is often overlooked when finding out problems for performance
issues with SQLServer
• Location
Separate tempdb from the user databases and SQL Server binaries and
place the tempdb on faster disk array.
alter database tempdb modify file(name=tempdev,
filename='D:\sqldata\tempdb.mdf')
alter database tempdb modify file(name=templog,
filename='D:\sqldata\templog.ldf')
CONFIDENTIAL
21-Jul-15
48
TempDB improvement
• DB Size
– Default: 8 MB
– Query the current size
SELECT name AS FileName,
size*1.0/128 AS FileSizeinMB,
CASE max_size
WHEN 0 THEN 'Autogrowth is off.'
WHEN -1 THEN 'Autogrowth is on.'
ELSE 'Log file will grow to a maximum size of 2 TB.'
END,
growth AS 'GrowthValue',
'GrowthIncrement' =
CASE
WHEN growth = 0 THEN 'Size is fixed and will not grow.'
WHEN growth > 0 AND is_percent_growth = 0
THEN 'Growth value is in 8-KB pages.'
ELSE 'Growth value is a percentage.'
END
FROM tempdb.sys.database_files;
CONFIDENTIAL
21-Jul-15
49
TempDB improvement
• Update the size
alter database tempdb modify file(name=tempdev,
filename='D:\sqldata\tempdb.mdf',
size=XXXMB)
CONFIDENTIAL
21-Jul-15
50
TempDB improvement
• Number of data files
– More data files for tempdb will reduce contention
– Recommend: number of processors in system.
alter database tempdb add file(name=tempdev2,
filename='d:\sqldata\tempdb2.ndf',
size=200MB,
filegrowth=10%)
• Separate data file and log file
– Tempdb does not need backups and it can well be in simple recovery
model. This is the recommendation and also the default.
CONFIDENTIAL
21-Jul-15
51
TempDB improvement
• Practice - Location
– Move C partition -> D partition
• Result
Page Response time (improvement in percent)
No Change
Move from
C: -> D:
with
default
size
Move from
C: -> D:
with
800M
size
Move from
C: -> D:
with
320M
size
Move from
C: -> D:
with
64M
size
Average
1.813051
1.434593
FAIL
1.701
1.64125
Maximum
5.983458
5.408847
FAIL
6.83278
6.060153
90 Percent
2.774966
2.236746
FAIL
2.48322
2.350949
CONFIDENTIAL
21-Jul-15
52
TempDB improvement (Resource
Compare)
CONFIDENTIAL
21-Jul-15
53
Production Data
CONFIDENTIAL
21-Jul-15
54
Production Data
Date
Release
Number
Total Action
Numbers
< 3s total
Average
action numbers Percent
Page <3s
5/15/2009
MKV 8.8
106,429
103,174
96.9%
5/13/2009
MKV 8.7.1
81,018
75,300
93%
4/20/2009
MKV 8.7
81,672
75,325
92.2%
CONFIDENTIAL
21-Jul-15
55
Production Data
Date
Release
Number
Total Action
Numbers
< 3s total
Average
action numbers Percent
Page <3s
5/15/2009
MKV 8.8
106,429
103,174
96.9%
5/13/2009
MKV 8.7.1
81,018
75,300
93%
4/20/2009
MKV 8.7
81,672
75,325
92.2%
CONFIDENTIAL
21-Jul-15
56
Some Hot Topics
• NUMA
• CLUSTERED DATABASE CONNECTION
CONFIDENTIAL
21-Jul-15
57
Some Hot Topics (NUMA
CONFIDENTIAL
21-Jul-15
58
Some Hot Topics (NUMA)
CONFIDENTIAL
21-Jul-15
59
Some Hot Topics (NUMA)
CONFIDENTIAL
21-Jul-15
60
Some Hot Topics (NUMA)
CONFIDENTIAL
21-Jul-15
61
Some Hot Topics (NUMA)
CONFIDENTIAL
21-Jul-15
62
Some Hot Topics (NUMA)
CONFIDENTIAL
21-Jul-15
63
Some Hot Topics (CLUSTERED
DATABASE CONNECTION)
CONFIDENTIAL
21-Jul-15
64
Some Hot Topics (CLUSTERED
DATABASE CONNECTION)
CONFIDENTIAL
21-Jul-15
65
Some Hot Topics (CLUSTERED
DATABASE CONNECTION)
CONFIDENTIAL
21-Jul-15
66
Some Hot Topics (CLUSTERED
DATABASE CONNECTION)
CONFIDENTIAL
21-Jul-15
67
Some Hot Topics (CLUSTERED
DATABASE CONNECTION)
• Microsoft JDBC support configure the failover partner information
dbs/mss/server = tstsapdb1;FailoverPartner=tstsapdb2;Database=TST
CONFIDENTIAL
21-Jul-15
68
Further topics
1. Disk tuning.
2. Sql server high availability and data replication
3. Sql server full text search
4. TempDB further tuning
CONFIDENTIAL
21-Jul-15
69
CONFIDENTIAL
21-Jul-15
70