How to use SQL Server T-SQL cursors

The general advice within the SQL Server community is do not use any T-SQL cursors at any time.
I must say that I have to agree, although if you have browsed my blog you've noticed that there are a couple of occasions when I do use cursors but only when the following is true:

- All cursors are read_only / forward-only for optimum speed.
- They're not used in OLTP environments.
- Datasources are small sets, usually system objects, like a list of databases / tables.
- The impact the cursor is many times smaller then the processing steps taken with the row results.
(like a cursor based on sys.databases and with the result of each row a full database backup is made).

Coding guidelines:
- Avoid using cursors and try solve the problem by using set based operations.
- Try to base them on readonly tables or local (#temptable) temporarily objects.
- Cleanup! Always CLOSE and DEALLOCATE cursors.
- Specify the 'LOCAL' option.
- Specify the FAST_FORWARD clause on DECLARE CURSOR. This opens an optimized forward-only, read-only cursor.
- Never use cursors in OLTP environments.
- Never use them as source of an update: UPDATE table SET col1 = 2 WHERE CURRENT OF the_cursor;
- Cursors are memory intensive; base them on smaller sets (less 5,000-10,000 rows, your site could differ: test!).

template:

    DECLARE the_cursor CURSOR FAST_FORWARD LOCAL FOR
        SELECT col1,col2,col3
         FROM dbo.mytable

    OPEN the_cursor

    FETCH NEXT FROM the_cursor INTO @col1,@col2,@col3
    WHILE (@@FETCH_STATUS <> -1)
    BEGIN

        FETCH NEXT FROM the_cursor INTO @col1,@col2,@col3
    END

    CLOSE the_cursor
    DEALLOCATE the_cursor
 

TIP:
If you do need cursors in high load OLTP environments because of some complex calculation that can't be done set based take the following approach:

Copy the required result set in a temporary object. Retrieve only the rows and columns you need, but do include all the fields of the primary key.

create #temptable (CalcValue int, pkCol int)

INSERT INTO #temptable (CalcValue, pkCol)
SELECT 0, PrimaryKeyCol
FROM dbo.HighLoadOLTPtable
WHERE -- your where clause here

Base your cursor on the temp. object.

Loop the cursor, perform your calculation and store the result in the temp. object row:
UPDATE #temptable SET CalcValue=complex_calculated_value
WHERE pkCol=pk_fields_as_fetched_by_cursor

When done looping close/deallocate the cursor, and update the high load OLTP source table(s) set based by primarykey, use a BEGIN TRANSACTION / COMMIT if required:

UPDATE dbo.HighLoadOLTPtable SET CalculatedValue = #temptable.CalcValueFROM dbo.HighLoadOLTPtable, #temptableWHERE dbo.HighLoadOLTPtable.PrimaryKeyCol = #temptable.pkCol

see my posts on TempDB configuration and TempDB objects

Working with SQL Server TempDB objects

Try to avoid them in the first place by using derived tables or CTEs

Table variables are easy an convenient, because you don't have to clean them up.
But only use them with small sets (smaller 1000 rows)

Do not use SELECT col1 INTO #temptable FROM dbo.table1

But do use a create table statement and a INSERT/SELECT:
CREATE #temptable (col1 int,col2 int)

INSERT INTO #temptable (col1,col2) SELECT Col1, Col2 FROM dbo.table1

a bit more work, but less blocking in environments with a higher load.

With larger sets you could consider using indexes and check with the query plan and statistics io to see if there are any benefits in using them.
Create these indexes after the table is filled with a set:

INSERT INTO #temptable (col1,col2) SELECT Col1,Col2 FROM dbo.table1
CREATE CLUSTERED INDEX CIX_#temptable ON #temptable (col1)
CREATE INDEX IX_#temptable_1 ON #temptable (Col2)

Create ony the indexes you absolutely need.

Configure TempDB as suggested in my post configure TempDB

Can your TempDB database disks handle the current load? See my post on fn_virtualfilestats

Prefer locally (#temptable) over globaly (##temptable) scoped definitions

Store results of linked servers in TempDB objects instead of accessing the remote datasource multiple times. See my post on linked servers

Be a good citizen, at the end of your stored procedure or T-SQL code always cleanup after your self and drop all temporarily objects you have created.

    --drop temp tables if they exists
    IF OBJECT_ID('tempdb..#temptable') IS NOT NULL
        DROP TABLE #temptable
 

Always run SQLIO on new servers!

When your vendor brings a new type of server on the market always check if the hardware is as good as they claim it is.

A while ago I did a test run and compared the IO results of a Dell server and an -almost- equal configured IBM server. Both were small configs with six internal drives. The tool I used was SQLIO.

SQLIO is a small .exe file (250KB) that mimics the SQL Server database engine. You don't have to install SQL Server for the tool to work. Only the Windows OS is enough, so you can also check the thoughput of your SAN filesystem.
You can download SQLIO and some documentation here

The Result? Guess what? The Dell server scored 10(ten!) times better on IO performance then the more expensive IBM...

Never take your vendors word for it, always test yourself. Because who gets the blaim in the end?

The test file was 25MB, but results were roughly the same for bigger test files.

Results for: sqlio -kR -s360 -frandom -o8 -b64 -LS -Fparam.txt

IBM server local drives:

throughput metrics:
IOs/sec: 1177.91
MBs/sec: 73.61
latency metrics:
Min_Latency(ms): 0
Avg_Latency(ms): 13
Max_Latency(ms): 919
histogram:
ms: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24+
%: 16 11 4 5 8 8 6 5 4 3 3 2 2 2 1 1 1 1 1 1 1 1 1 1 13

Dell server local drives:

throughput metrics:
IOs/sec: 11,292.99
MBs/sec: 705.81
latency metrics:
Min_Latency(ms): 0
Avg_Latency(ms): 1
Max_Latency(ms): 394
histogram:
ms: 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24+
%: 0 96 4 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0

See my post on disk configuration tips.

SQL Server database settings

When creating a new SQL Server database I always use this checklist:

Name: Use as less non-alpha numeric characters in the database name as possible (do not use the ‘-‘, it will mess up your scripts). Use pascal casing for readability (use OrdersDb, do not use Orders-db or Orders_db)

Set Page verify to Checksum

Pre allocate data and log size to a reasonable size, do not rely on autogrowth.
Set a reasonable Autogrow size in MB (not a percentage) for data and log files

Enabling auto create/update statistics, is fine for most situations.

Set the right recovery model.
If your using ‘full’ or ‘bulk recovery’, schedule regular log backups (see this post )

Try to create at least two –equally sized!- device files for the data devices.
Not only reduces this locking issues but also when a database is regular under a heavy load,files can easily be moved to additional physical volumes (see this post )

Create one file for the Log

Use fn_virtualfilestats() to measure I/O load per device file (see this post )

Set the .MDF, .NDF and .LDF files in their own folders with the same name as the database. Don’t store all your db’s in one folder

Reorganize indexes/statistics on a regular basis (see this post )

Run DBCC CHECKDB on a regular basis (see this post )

Do not schedule any shrink operations

Configure instant file initialization (see this post )

Check out my post on auto index management and tools for performance, to setup more databases on the same instance and save costs

SQL Server disk configuration tips

Most important metrics for SQL Server: ‘throughput’: how many MB/s can the disks handle and what is the load placed on the disks.

If you’re buying a standard package your vendor should know this.
Surprisingly however in the hardware proposal (top end server(s) / high rpm disks right?) this figure isn’t mentioned anywhere in most situations.
What you can do, is ask your vendor if they could run a fn_virualfilestats() function (see my fn_virtualfilestats post) on a similar production configuration of another customer (better: talk to them yourself!) . This gives you insight on the I/O pattern (more read of more write activity) and the load on TempDB, to determine if it’s a good candidate –which it usually is- to store on it’s own disk spindles.

Run the Microsoft SQLIO tool to determine a baseline on MB/s and latency per disk partition.
Store it for later reference, the Windows performance tool can also measure the throughput and it gives you the possibility to compare the numbers and check if there might be a disk throughput performance bottleneck in your production system.

Disk configuration tips:

Use as many disk spindles as you can, and try spreading data / log and tempdb on different raid arrays.
Stripe your RAID-config at 256k instead of 64k / 128k.This will increase performance of SQL Server read-aheads.
If you use Windows 2000 / 2003 align the disk partitions!!!:
(see Miscrosoft knowledge base article KB300415 on how to use Diskpart).
At the Diskpart command prompt,
type: Create Partition Primary Align=X,
where X is either 32 or 64, depending on the recommendation from your storage vendor.
If your storage vendor does not have any specific recommendations, it is recommended that
you use 64.
This step is not needed in Windows 2008

Full (not quick) format with a 64kb stripe size for disks holding data/log and backup files.

Full (not quick) format with a 4kb stripe size for OS and program/swap devices

Data raid5 or raid10 when there’s more write then read activity (check with fn_virualfilestats),
Log (raid1 or raid10 under heavy load), separate backup disk (raid5)

Put TempDB on its own physical disks (raid1, raid10 under heavy load).
In a non-clustered environment, I normally use the local server storage for TempDB.

-->
>
Use NTFS volumes

Do not use more then 80% of the volume

Defragment the disks on a regular basis

How to configure SQL Server TempDB

As a general guideline, create one data file for each CPU on the server.
Note that a dual-core CPU is considered to be two CPUs. logical procs (hyperthreading) do not

Only create one Log file

Do not use autoshrink on TempDB

If your'e using SQL Server 2008 set Page verify to ChecksumInvestigate the possibility if you can switch off the properties ‘Auto create’ and ‘Auto Update’ statistics,it can speed up the creation of objects in TempDB. Be careful however: measure the impact these changes can have on production! Change and measure only one parameter at the time.

After installation move the TempDB database to its own disks that differ from those that are used by user databases

ALTER DATABASE TempDB
MODIFY FILE (NAME=tempdev, FILENAME= '\tempdb.mdf');
GO

ALTER DATABASE TempDB
MODIFY FILE (NAME=templog, FILENAME= '\TempLog.ldf');
GO

Restart SQL Server service (not the Windows server), to make changes permanent

Pre-allocate data and log device sizes (create files with same equal size), do not rely on small auto growth steps

If you use auto growth set it to a reasonable size in MB, not a percentage

TempDB file size default FILEGROWTH increment
0 to 100 MB growth:10 MB
100 to 200 MB growth:20 MB
200 to 1000 MB growth:50 to 75 MB
1 GB or More growth:150 to 250 MB

Measure TempDB usage over time with fn_virtualfilestats
New releases of your –vendor- software could have a different impact on TempDB load.

Do you have a datetime T-SQL 'BETWEEN AND' bug??

Create a table with a datetime column

create table #t (col1 datetime)

add couple of rows with date and time values

insert into #t values ('1/mar/2009 08:00')
insert into #t values ('1/mar/2009 09:00')
insert into #t values ('1/mar/2009 10:00')

get the rows

select * from #t
where col1 between '1/mar/2009' and '1/mar/2009'

and guess what?
(0 row(s) affected)


Why? You’re trying to select a date without a time and that's treated as 00:00 midnight
and these rows do not exist in the table

Three ways to solve:

1.
append a time value to the sql string:

select * from #t
where col1 between '1/mar/2009 00:00:00' and '1/mar/2009 23:59:59'
(3 row(s) affected)

no go, too much work -specially with variables- and could lead to (more) bugs


2.
select * from #t
where CONVERT(datetime,CONVERT(varchar(20), col1, 112),112) between '1/mar/2009' and '1/mar/2009'
(3 row(s) affected)

uses double convert() T-SQL function to get rid of the time part and switch back to datetime format

no go: if you have a index defined an the column it will not be used by the SQL Server query optimizer because of the convert function. Bad performance!

3.
my absolute thumbs up favorite:

select * from #t
where col1 >= '1/mar/2009' and col1 < dateadd(d,1,'1/mar/2009')
(3 row(s) affected)

GO, this construction has served me well over the years: date/time values are handled correctly,
indexes are used and no hassle with convert() leading to good performance and good maintainable T-SQL code!

have fun

SQL Server tools for performance

Since the introduction of SQL Server 2005 Microsoft implemented Dynamic Management Views (DMV’s) in the database engine that enable us to peek inside of it, and retrieve the internal –performance- counters.

In this post I present a couple of queries gathered and used over time that assisted me to pin down the problem when having performance related issues.

If you’ve inherited a system from an older / upgraded SQL Server version, I recommend running the ‘unused indexes‘ query mentioned later on. It saved my life in more than one occasion.
In my experience as tables have grown bigger, indexes were created for specific needs and over time became obsolete, but no one bothered to remove them, leaving the system with a lot of unnecessary I/O overhead.

Explore sys.dm_exec_query_stats it contains more fields than used here (last_physical_reads for instance) and also check out other DMVs for additional info you might be interested in.

Before starting any investigation make sure that all indexes are defragmented and all the statistics are up to date up front (see my post on index reorganization)

The queries described here, combined with my previous posts on fn_virtualfilestats, auto index management, STATISTICS IO, multiple files per database and index reorganization should supply you with enough info to tackle most of I/O related performance issues.

Beware: DMV’s are reset when the SQL Server instance is restarted! So the numbers are only meaningful when the db engine is running for a considerable amount of time

First determine how long the instance has been running:

SELECT crdate AS Instance_Started_On
FROM sysdatabases
WHERE name = 'tempdb'
 

Queries with highest IO load:

SELECT TOP 100
        total_logical_reads, total_logical_writes, execution_count,
        total_logical_reads+total_logical_writes AS total_IO,
        st.TEXT AS query_text,
    st.dbid AS database_id,
        DB_NAME(st.dbid) AS database_name
    FROM sys.dm_exec_query_stats  qs
   CROSS APPLY sys.dm_exec_sql_text(sql_handle) st
   WHERE total_logical_reads+total_logical_writes > 0
   ORDER BY (total_logical_reads+total_logical_writes) DESC -– most i/o intensive
 

Queries using the most CPU:

SELECT TOP 100
    highest_cpu_queries.total_worker_time,
    sql_text.dbid,
    DB_NAME(sql_text.dbid) AS database_name,
    sql_text.[TEXT]
FROM  
    (SELECT TOP 100 PERCENT
        qs.plan_handle,  
        qs.total_worker_time
    FROM  
        sys.dm_exec_query_stats qs
    ORDER BY qs.total_worker_time DESC) AS highest_cpu_queries
    cross apply sys.dm_exec_sql_text(plan_handle) AS sql_text
ORDER BY highest_cpu_queries.total_worker_time DESC
 

Get the unused indexes:

USE your_db_name –- switch TO your own db here BEFORE executing the dmv query below
go

SELECT TOP 100
OBJECT_NAME(i.OBJECT_ID) AS tablename,
i.name AS indexname,
s.user_updates,
s.user_seeks + s.user_scans + s.user_lookups AS totalreads,
s.user_updates AS totalwrites
FROM sys.indexes i
            LEFT join sys.dm_db_index_usage_stats s
ON s.OBJECT_ID = i.OBJECT_ID and  
                  i.index_id = s.index_id
WHERE OBJECTPROPERTY(i.OBJECT_ID, 'IsIndexable') = 1 and
-- index_usage_stats has no reference to this index (not being used)
s.index_id IS null or
-- index is being updated, but not used by seeks/scans/lookups
(s.user_updates > 0 and s.user_seeks = 0 and s.user_scans = 0 and s.user_lookups = 0)
ORDER BY s.user_updates DESC
go
 

SET STATISTICS IO ON!

The graphical query plans introduced in the latest SQL Server versions give you great inside info on how your query is executing. Downside: sometimes they are difficult to read and they take time to fully understand.

Enter: SET STATISTICS IO ON

It gives you an overview off how many I/Os your query is executing on each accessed table, and by working together with the query optimizer you can absolutely minimize these numbers.
And by doing so optimize and increase the work load a SQL Server instance can handle.

For example: what is the impact of replacing an SELECT IN sub-query by an IF EXISTS () construction.

This feature is available since SQL Server 6.x, so you can still tune your old instances.

In my opinion this option should always be used with any new or changed query against a representative amount of data: a copy from, or simulation of, a production environment to check if the results are in-line with expectations.

Here’s a simple query and result when the option is enabled in the message tab of the results pane (click to enlarge):










If you have a hosted SQL Server environment with tight security, it could well be that generating graphical query plans is prohibited, while this option could still be granted giving you lots of meaningful information.

Before diving in to it check that all indexes a reorganized and the statistics are up to date.
(see my post index defragmentation)

Also check my post auto index management and let SQL Server assist you with your indexes!

Oh yeah, switch it to off again when you are done tuning your queries, don’t include it in your stored procedures.

Create multiple files per database and increase performance

In this post I want to explain why to create multiple data files per SQL Server database

Note: this is only relevant for data files, not for log files. Log files always have one file!

Microsoft documentation says:
• The number of data files within a single filegroup should equal to the number of physical CPU cores (hyper threading cores should not be counted)

This is also true for TempDB

So create all the files in your new database with the same initial size – make an estimate on how big you want the files to be- and set a reasonable growth factor. Do not set it to a percentage, but on the same fixed size in MB for all data files. Use the same growth factor for all the files you have created.

SQL Server will fill the data files in a round robin way, meaning that data is spread equally across all the files as the database grows.

Tip!
Another advantage from this approach is that if you have performance issues and you can pin this down to an disk I/O related problem
(see my http://sqlserveradvisor.blogspot.com/2009/02/microsoft-included-fnvirtualfilestats.html on fn_virtualfilestats() post for this), it is then an easy change to add a new raid disk configuration to your server and then spread the data files evenly over both the raid arrays.
Giving you an instant performance boost!!

fn_virtualfilestats: Measure the I/O load on your disk system

Microsoft included the fn_virtualfilestats function in SQL Server so you can monitor the I/O load the instance executes on the disk subsystem.

I use this function mainly to:
- Get the load on TempDB and check if it’s a good idea to move it to it’s own disks (if you have not already done so)
- Get the I/O profile (more read of write activity?) on the database level to help determine if it might be an idea to replace a raid-5 array to a raid-10 or to move datafiles to additional diskarrays (see my post http:// for this)
- Check if there are any disk related issues, in other words are there many stalled I/O’s?
(stalled I/O is an I/O waiting for another I/O)
- Check if the system can handle more db’s, eliminating additional servers

Real life example:
I managed to save big $$$ and could avoid buying new hardware by letting our vendor run this script in comaparable environent of another customer and so proving the load was not so big as they claimed, so we could run the new system on our existing hardware.

Here’s the T-SQL to get the I/O profile of all your database and data files:

DECLARE @TotalIO    BIGINT,
    @TotalBytes BIGINT,
    @TotalStall BIGINT

SELECT @TotalIO  = SUM(NumberReads + NumberWrites),
       @TotalBytes = SUM(BytesRead + BytesWritten),
       @TotalStall = SUM(IoStallMS)
FROM ::FN_VIRTUALFILESTATS(NULL, NULL)

SELECT [DbName] = DB_NAME([DbId]),
      (SELECT name FROM sys.master_files
        WHERE database_id = [DbId]
              and FILE_ID = [FileId]) filename,
    [%ReadWrites]       = (100 * (NumberReads + NumberWrites) / @TotalIO),
    [%Bytes]        = (100 * (BytesRead + BytesWritten) / @TotalBytes),
    [%Stall]        = (100 * IoStallMS / @TotalStall),
    [NumberReads],
    [NumberWrites],
    [TotalIO]       = CAST((NumberReads + NumberWrites) AS BIGINT),
    [MBsRead]       = [BytesRead] / (1024*1024),
    [MBsWritten]        = [BytesWritten] / (1024*1024),
    [TotalMBs]      = (BytesRead + BytesWritten) / (1024*1024),
    [IoStallMS],
    IoStallReadMS,
    IoStallWriteMS,
    [AvgStallPerIO]     = ([IoStallMS] / ([NumberReads] + [NumberWrites] + 1)),
    [AvgStallPerReadIO] = (IoStallReadMS / ([NumberReads] + 1)),
    [AvgStallPerWriteIO]= (IoStallWriteMS / ( [NumberWrites] + 1)),

    [AvgBytesPerRead]  = ((BytesRead) / (NumberReads + 1)),
    [AvgBytesPerWrite] = ((BytesWritten) / (NumberWrites + 1))
FROM ::FN_VIRTUALFILESTATS(NULL, NULL)
ORDER BY dbname
 

Be ware: counters are re-set after the instance is restarted!

The result (not all columns shown here). Click on image to enlarge

Let SQL Server automatically suggest indexes!

I found this link on the Microsoft website
http://blogs.msdn.com/queryoptteam/archive/2006/06/01/613516.aspx
and ran the AutoIndex.sql from the attached AutoIndexBlog.zip in my code own CodeCatalog db

The two sp’s I use are ‘auto_create_index’ and ‘auto_drop_index’. Making use of the dynamic management views: dm_db_index_operational_stats, sys.dm_db_index_usage_stats and sys.dm_db_index_physical_stats all databases/tables are analyzed. The result with the suggested create and drop index statements are stored in the ‘recommendations’ table

I schedule the two mentioned sp’s in one SQL Server Agent job every night. The performance impact is low, they’re completed in less then a minute.

Let the sp’s run on a daily basis and monitor the suggestions over time.
It happened to me that indexes where suggested and later one dropped again,
usually when –new- tables change over time.


Results as found in the recommendations table (click to enlarge):

It’s a really great source when having discussions with vendors and finally pinpoint what they are doing wrong!