T-SQL Query Performance Tuning Tips

In this article, I will discuss some useful T-SQL query performance tips and tricks for SQL server developers.

In this article, I will discuss some T-SQL query performance tips and tricks for SQL Server programmers. The tips mentioned in this article may sound obvious to most of you, but I have seen professional developers who don't think before using them.

My first tip is to not a WHERE clause in your SELECT statement to narrow the number of rows returned. If you don't use a WHERE clause, then SQL Server will perform a table scan of your table and return all of the rows. In some cases, you may want to return all rows, and not using a WHERE clause is appropriate in this case. But if you don't need all the rows returned then use a WHERE clause to limit the number of rows returned.
 
By returning data you don't need, you are causing SQL Server to perform I/O it doesn't need to perform, wasting SQL Server resources. In addition, it increases network traffic, that can also lead to reduced performance. And if the table is very large, a table scan will lock the table during the time-consuming scan, preventing other users from accessing it, hurting concurrency.

Another negative aspect of a table scan is that it will tend to flush out data pages from the cache with useless data, that reduces SQL Server's ability to reuse useful data in the cache, that increases disk I/O and hurts performance. [6.5, 7.0, 2000]

To help identify long running queries, use the SQL Server Profiler Create Trace Wizard to run the "TSQL By Duration" trace. You can specify the length of the long running queries you are trying to identify (such as over 1000 milliseconds), and then have these recorded in a log for you to investigate later. [7.0]

When using the UNION statement, keep in mind that, by default, it performs the equivalent of a SELECT DISTINCT on the final result set. In other words, UNION takes the results of two like recordsets, combines them, and then performs a SELECT DISTINCT in order to eliminate any duplicate rows. This process occurs even if there are no duplicate records in the final recordset. If you know that there are duplicate records, and this presents a problem for your application then by all means use the UNION statement to eliminate the duplicate rows.

On the other hand, if you know that there will never be any duplicate rows, or if there are, and this presents no problem to your application, then you should use the UNION ALL statement instead of the UNION statement. The advantage of the UNION ALL is that is does not perform the SELECT DISTINCT function, that saves a lot of unnecessary SQL Server resources from being used. [6.5, 7.0, 2000]

Sometimes you might want to merge two or more sets of data resulting from two or more queries using UNION. For example:

  1. SELECT column_name1, column_name2  
  2. FROM table_name1  
  3. WHERE column_name1 = some_value  
  4. UNION  
  5. SELECT column_name1, column_name2  
  6. FROM table_name1  
  7. WHERE column_name2 = some_value  

This same query can be rewritten, as in the following example, and when doing so, performance will be increased:

  1. SELECT DISTINCT column_name1, column_name2  
  2. FROM table_name1  
  3. WHERE column_name1 = some_value OR column_name2 = some_value  
And if you can assume that neither criteria will return duplicate rows then you can even further boost the performance of this query by removing the DISTINCT clause. [6.5, 7.0, 2000]

Carefully evaluate whether your SELECT query needs a DISTINCT clause or not. Some developers automatically add this clause to every one of their SELECT statements, even when it is not necessary. This is a bad habit that should be stopped.

The DISTINCT clause should only be used in SELECT statements if you know that duplicate returned rows are a possibility, and that having duplicate rows in the result set would cause problems with your application.

The DISTINCT clause creates a lot of extra work for SQL Server, and reduces the physical resources that other SQL statements have at their disposal. Because of this, only use the DISTINCT clause if it is necessary. [6.5, 7.0, 2000]

In your queries, don't return column data you don't need. For example, you should not use SELECT * to return all the columns from a table if you don't need all the data from each column. In addition, using SELECT * prevents the use of covered indexes, further potentially hurting query performance. [6.5, 7.0, 2000]

If your application allows users to run queries, but you are unable in your application to easily prevent users from returning hundreds, even thousands of unnecessary rows of data they don't need, consider using the TOP operator within the SELECT statement. This way, you can limit how many rows are returned, even if the user doesn't enter any criteria to help reduce the number of rows returned to the client. For example, the statement: 
  1. SELECT TOP 100 fname, lname FROM customers  
  2. WHERE state = 'AP'  
Limits the results to the first 100 rows returned, even if 10,000 rows actually meet the criteria of the WHERE clause. When the specified number of rows is reached, all processing on the query stops, potentially saving SQL Server overhead, and boosting performance.

The TOP operator works by allowing you to specify a specific number of rows to be returned, like the example above, or by specifying a percentage value, like this:
  1. SELECT TOP 10 PERCENT fname, lname FROM customers  
  2. WHERE state = 'AP'  

In the preceding example, only 10 percent of the available rows would be returned. Keep in mind that using this option may prevent the user from getting the data they need. For example, the data they are looking for may be in record 101, but they only get to see the first 100 records. Because of this, use this option with discretion. [7.0, 2000]

You may have heard of a SET command called SET ROWCOUNT. Like the TOP operator, it is designed to limit how many rows are returned from a SELECT statement. The SET ROWCOUNT and the TOP operator effectively perform the same function.
While is most cases, using either option works equally efficiently, there are some instances (such as rows returned from an unsorted heap) where the TOP operator is more efficient than using SET ROWCOUNT. Because of this, using the TOP operator is preferable to using SET ROWCOUNT to limit the number of rows returned by a query. [6.5, 7.0, 2000]

In a WHERE clause, the various operators used directly affect how fast a query is run. This is because some operators lend themselves to speed over other operators. Of course, you may not have any choice of which operator you use in your WHERE clauses, but sometimes you do.

Here are the key operators used in the WHERE clause, ordered by their performance. Those operators at the top will produce results faster than those listed at the bottom.

  • =
  • >, >=, <, <=
  • LIKE
  • <>

The lesson here is to use = as much as possible, and <> as least as possible. [6.5, 7.0, 2000]

In a WHERE clause, the various operands used directly affect how fast a query is run. This is because some operands lend themselves to speed over other operands. Of course, you may not have any choice of which operand you use in your WHERE clauses, but sometimes you do.

Here are the key operands used in the WHERE clause, ordered by their performance. Those operands at the top will produce results faster than those listed at the bottom.

  • A single literal used by itself on one side of an operator
  • A single column name used by itself on one side of an operator, a single parameter used by itself on one side of an operator
  • A multi-operand expression on one side of an operator
  • A single exact number on one side of an operator
  • Other numeric numbers (other than exact), date and time
  • Character data, NULLs

The simpler the operand, and using exact numbers, provides the best overall performance. [6.5, 7.0, 2000]

If a WHERE clause includes multiple expressions, there is generally no performance benefit gained by ordering the various expressions in any particular order. This is because the SQL Server Query Optimizer does this for you, saving you the effort. There are a few exceptions to this, which are discussed on this web site. [7.0, 2000]

Don't include code that doesn't do anything. This may sound obvious, but I have seen this in some off-the-shelf SQL Server-based applications. For example, you may see code like this:

  1. SELECT column_name FROM table_name  
  2. WHERE 1 = 0  
When this query is run, no rows will be returned. Obviously, this is a simple example (and most of the cases where I have seen this done have been very long queries), a query like this (or part of a larger query) like this doesn't perform anything useful, and shouldn't be run. It is just wasting SQL Server resources. In addition, I have seen more than one case where such dead code actually causes SQL Server to throw errors, preventing the code from even running. [6.5, 7.0, 2000] 

By default, some developers, especially those who have not worked with SQL Server before, routinely include code similar to this in their WHERE clauses when they make string comparisons:

  1. SELECT column_name FROM table_name  
  2. WHERE LOWER(column_name) = 'name'  
In other words, these developers are making the assumption that the data in SQL Server is case-sensitive, which it generally is not. If your SQL Server database is not configured to be case sensitive then you don't need to use LOWER or UPPER to force the case of text to be equal for a comparison to be performed. Just leave these functions out of your code. This will speed up the performance of your query, as any use of text functions in a WHERE clause hurts performance.

But what if your database has been configured to be case-sensitive? Should you then use the LOWER and UPPER functions to ensure that comparisons are properly compared? No. The preceding example is still poor coding. If you need to deal with ensuring case is consistent for proper comparisons, use the technique described below, along with appropriate indexes on the column in question:
  1. SELECT column_name FROM table_name  
  2. WHERE column_name = 'NAME' or column_name = 'name'  
This code will run much faster than the first example. [6.5, 7.0, 2000]

Try to avoid WHERE clauses that are non-sargable. The term "sargable" (which is in effect a made-up word) comes from the pseudo-acronym "sarg", which stands for "search argument," that refers to a WHERE clause that compares a column to a constant value. If a WHERE clause is sargable, this means that it can take advantage of an index (assuming one is available) to speed completion of the query. If a WHERE clause is non-sargable, this means that the WHERE clause (or at least part of it) cannot take advantage of an index, instead performing a table/index scan, that may cause the query's performance to suffer.
 
Non-sargable search arguments in the WHERE clause, such as "IS NULL", "<>", "!=", "!>", "!<", "NOT", "NOT EXISTS", "NOT IN", "NOT LIKE", and "LIKE '%500'" generally prevents (but not always) the query optimizer from using an index to perform a search. In addition, expressions that include a function on a column, expressions that have the same column on both sides of the operator, or comparisons against a column (not a constant), are not sargable.

But not every WHERE clause that has a non-sargable expression in it is doomed to a table/index scan. If the WHERE clause includes both sargable and non-sargable clauses, then at least the sargable clauses can use an index (if one exists) to help access the data quickly.

In many cases, if there is a covering index on the table that includes all of the columns in the SELECT, JOIN, and WHERE clauses in a query, then the covering index can be used instead of a table/index scan to return a query's data, even if it has a non-sargable WHERE clause. But keep in mind that covering indexes have their own drawbacks, such as producing very wide indexes that increase disk I/O when they are read.

In some cases, it may be possible to rewrite a non-sargable WHERE clause into one that is sargable. For example, the clause:
  1. WHERE SUBSTRING(firstname,1,1) = 'm'  
Can be rewritten like this:
  1. WHERE firstname like 'm%  
Both of these WHERE clauses produce the same result, but the first one is non-sargable (it uses a function) and will run slow, while the second one is sargable, and will run much faster. 

WHERE clauses that perform some function on a column are non-sargable. On the other hand, if you can rewrite the WHERE clause so that the column and function are separate, then the query can use an available index, greatly boosting performance. For example, in the following example, a function acts directly on a column, and the index cannot be used:
  1. SELECT member_number, first_name, last_name  
  2. FROM members  
  3. WHERE DATEDIFF(yy,datofbirth,GETDATE()) > 21  
In the following example, a function has been separated from the column and an index can be used:
  1. SELECT member_number, first_name, last_name  
  2. FROM members  
  3. WHERE dateofbirth < DATEADD(yy,-21,GETDATE())  
Each of the preceding queries produces the same results, but the second query will use an index because the function is not performed directly on the column, as it is in the first example. The moral of this story is to try to rewrite WHERE clauses that have functions so that the function does not act directly on the column.

WHERE clauses that use NOT are not sargable, but can often be rewritten to remove the NOT from the WHERE clause, for example:
  1. WHERE NOT column_name > 5  
To
  1. WHERE column_name <= 5  
Each of the preceding clauses produces the same results, but the second one is sargable. If you don't know if a particular WHERE clause is sargable or non-sargable, check out the query's execution plan in Query Analyzer. Doing this, you can very quickly see if the query will be using index lookups or table/index scans to return your results. 

With some careful analysis, and some clever thought, many non-sargable queries can be written so that they are sargable. Your goal for best performance (assuming it is possible) is to get the left side of a search condition to be a single column name, and the right side an easy to look up the value. [6.5, 7.0, 2000]

If you run into a situation where a WHERE clause is not sargable because of the use of a function on the right side of an equality sign (and there is no other way to rewrite the WHERE clause) then consider creating an index on a computed column instead. This way, you avoid the non-sargable WHERE clause altogether, using the results of the function in your WHERE clause instead. Because of the additional overhead required for indexes on computed columns, you will only want to do this if you need to run this same query over and over in your application, thereby justifying the overhead of the indexed computed column. [2000]

If you currently have a query that uses NOT IN then that offers poor performance because the SQL Server optimizer has to use a nested table scan to perform this activity, instead, try to use one of the following options instead, all of which offer better performance:

  • Use EXISTS or NOT EXISTS
  • Use IN
  • Perform a LEFT OUTER JOIN and check for a NULL condition
    [6.5, 7.0, 2000]

When you have a choice of using the IN or the EXISTS clause in your Transact-SQL then you will generally want to use the EXISTS clause, since it is usually more efficient and performs faster. [6.5, 7.0, 2000]

If you find that SQL Server uses a TABLE SCAN instead of an INDEX SEEK when you use an IN or OR clause as part of your WHERE clause, even when those columns are covered by an index then consider using an index hint to force the Query Optimizer to use the index.

For example:

  1. SELECT * FROM tblTaskProcesses WHERE nextprocess = 1 AND processid IN (8,32,45)  
Takes about 3 seconds, whereas:
  1. SELECT * FROM tblTaskProcesses (INDEX = IX_ProcessID) WHERE nextprocess = 1 AND processid IN (8,32,45)  
returns in under a second. [7.0, 2000]

If you use LIKE in your WHERE clause then try to use one or more leading characters in the clause, if at all possible. For example, use:

  1. LIKE 'm%'

not:

  1. LIKE '%m'   
If you use a leading character in your LIKE clause, then the Query Optimizer has the ability to potentially use an index to perform the query, speeding performance and reducing the load on SQL Server.

But if the leading character in a LIKE clause is a wildcard then the Query Optimizer will not be able to use an index, and a table scan must be run, reducing performance and taking more time.

The more leading characters you can use in the LIKE clause, the more likely the Query Optimizer will find and use a suitable index. [6.5, 7.0, 2000]

If your application needs to retrieve summary data often, but you don't want to have the overhead of calculating it on the fly every time it is needed, consider using a trigger that updates summary values after each transaction into a summary table.

While the trigger has some overhead, overall, it may be less than having to calculate the data every time the summary data is needed. You may need to experiment to see which methods is fastest for your environment. [6.5, 7.0, 2000]

If your application needs to insert a large binary value into an image data column, perform this task using a Stored Procedure, not using an INSERT statement embedded in your application.

The reason for this is because the application must first convert the binary value into a character string (that doubles its size, thus increasing network traffic and taking more time) before it can be sent to the server. And when the server receives the character string, it then must convert it back to the binary format (taking even more time).

Using a Stored Procedure avoids all this because all the activity occurs on the SQL Server, and little data is transmitted over the network. [6.5, 7.0, 2000]

When you have a choice of using the IN or the BETWEEN clauses in your Transact-SQL, you will generally want to use the BETWEEN clause, since it is much more efficient. For example:
  1. SELECT customer_number, customer_name  
  2. FROM customer  
  3. WHERE customer_number in (1000, 1001, 1002, 1003, 1004)  
Is much less efficient than this:
  1. SELECT customer_number, customer_name  
  2. FROM customer  
  3. WHERE customer_number BETWEEN 1000 and 1004  
Assuming there is a useful index on customer_number, the Query Optimizer can locate a range of numbers much faster (using BETWEEN) than it can find a series of numbers using the IN clause (that is really just another form of the OR clause). [6.5, 7.0, 2000]

If possible, try to avoid using the SUBSTRING function in your WHERE clauses. Depending on how it is constructed, using the SUBSTRING function can force a table scan instead of allowing the optimizer to use an index (assuming there is one). If the substring you are searching for does not include the first character of the column you are searching for, then a table scan is performed.
 
If possible, you should avoid using the SUBSTRING function and use the LIKE condition instead, for better performance.

Instead of doing this:
  1. WHERE SUBSTRING(column_name,1,1) = 'b'  
Try using this instead: 
  1. WHERE column_name LIKE 'b%'  
If you decide to make this choice, keep in mind that you will want your LIKE condition to be sargable, that means that you cannot place a wildcard in the first position. [6.5, 7.0, 2000]

Where possible, avoid string concatenation in your Transact-SQL code, as it is not a fast process, contributing to overall slower performance of your application. [6.5, 7.0, 2000]

Generally, avoid using optimizer hints in your queries. This is because it is generally very hard to outguess the Query Optimizer. Optimizer hints are special keywords that you include with your query to force how the Query Optimizer runs. If you decide to include a hint in a query, this forces the Query Optimizer to become static, preventing the Query Optimizer from dynamically adapting to the current environment for the given query. More often than not, this hurts, not helps performance.

If you think that a hint might be necessary to optimize your query then be sure you first do all of the following first:
  • Update the statistics on the relevant tables.
  • If the problem query is inside a Stored Procedure, recompile it.
  • Review the search arguments to see if they are sargable, and if not, try to rewrite them so that they are sargable.
  • Review the current indexes, and make changes if necessary.

If you have done all of the preceding, and the query is not running as you expect, then you may want to consider using an appropriate optimizer hint. If you haven't heeded my advice and have decided to use some hints, keep in mind that, as your data changes, and as the Query Optimizer changes (through service packs and new releases of SQL Server), your hard-coded hints may no longer offer the benefits they once did. So if you use hints then you need to periodically review them to see if they are still performing as expected. [6.5, 7.0, 2000]

If you have a WHERE clause that includes expressions connected by two or more AND operators, SQL Server will evaluate them from left to right in the order they are written. This assumes that no parenthesis have been used to change the order of execution. Because of this, you may want to consider one of the following when using AND:

  • Locate the least likely true AND expression first. This way, if the AND expression is false then the clause will end immediately, saving time.
  • If both parts of an AND expression are equally likely to be false then put the least complex AND expression first. This way, if it is false, less work will need to be done to evaluate the expression.

You may want to consider using a Query Analyzer to look at the execution plans of your queries to see which is best for your situation. [6.5, 7.0, 2000]

If you want to boost the performance of a query that includes an AND operator in the WHERE clause then consider the following:

  • Of the search criterions in the WHERE clause, at least one of them should be based on a highly selective column that has an index.
  • If at least one of the search criterions in the WHERE clause is not highly selective then consider adding indexes to all of the columns referenced in the WHERE clause.
  • If none of the columns in the WHERE clause are selective enough to use an index on their own, consider creating a covering index for this query.
    [7.0, 2000]

The Query Optimizer will perform a table scan or a clustered index scan on a table if the WHERE clause in the query contains an OR operator and if any of the referenced columns in the OR clause are not indexed (or does not have a useful index).

Because of this, if you use many queries with OR clauses then you will want to ensure that each referenced column in the WHERE clause has a useful index. [7.0, 2000]

A query with one or more OR clauses can sometimes be rewritten as a series of queries that are combined with a UNION ALL statement, in order to boost the performance of the query. For example, let's have a look at the following query:

  1. SELECT employeeID, firstname, lastname  
  2. FROM names  
  3. WHERE dept = 'prod' or city = 'Orlando' or division = 'food'  
This query has three separate conditions in the WHERE clause. In order for this query to use an index, then there must be an index on all three columns found in the WHERE clause.

This same query can be written using UNION ALL instead of OR, as in this example: 
  1. SELECT employeeID, firstname, lastname FROM names WHERE dept = 'prod'  
  2. UNION ALL  
  3. SELECT employeeID, firstname, lastname FROM names WHERE city = 'Orlando'  
  4. UNION ALL  
  5. SELECT employeeID, firstname, lastname FROM names WHERE division = 'food'  
Each of these queries will produce the same results. If there is only an index on dept, but not the other columns in the WHERE clause, then the first version will not use any index and a table scan must be performed. But in the second version of the query will use the index for part of the query, but not for all of the query.

Admittedly, this is a very simple example, but even so, it does demonstrate how rewriting a query can affect whether or not an index is used or not. If this query was much more complex, then the approach of using UNION ALL might be must more efficient, since it allows you to tune each part of the index individually, something that cannot be done if you use only ORs in your query.

Note that I am using UNION ALL instead of UNION. The reason for this is to prevent the UNION statement from trying to sort the data and remove duplicates, which hurts performance. Of course, if there is the possibility of duplicates, and you want to remove them, then of course you can use just UNION.

If you have a query that uses ORs that do not make the best use of indexes then consider rewriting it as a UNION ALL, and then testing the performance. Only through testing can you be sure that one version of your query will be faster than another. [7.0, 2000]

Don't use ORDER BY in your SELECT statements unless you really need to, since it adds a lot of extra overhead. For example, perhaps it may be more efficient to sort the data at the client than at the server. In other cases, perhaps the client doesn't even need sorted data to achieve its goal. The key here is to remember that you shouldn't automatically sort data, unless you know it is necessary. [6.5, 7.0, 2000]

Whenever SQL Server needs to perform a sorting operation, additional resources need to be used to perform this task. Sorting often occurs when any of the following Transact-SQL statements are executed:

  • ORDER BY
  • GROUP BY
  • SELECT DISTINCT
  • UNION
  • CREATE INDEX (generally not as critical as happens much less often)

In many cases, these commands cannot be avoided. On the other hand, there are few ways that sorting overhead can be reduced. These include:

  • Keep the number of rows to be sorted to a minimum. Do this by only returning those rows that absolutely need to be sorted.
  • Keep the number of columns to be sorted to the minimum. In other words, don't sort more columns than required.
  • Keep the width (physical size) of the columns to be sorted to a minimum.
  • Sort a column with number datatypes instead of character datatypes.

When using any of the preceding Transact-SQL commands, try to keep the preceding performance-boosting suggestions in mind. [6.5, 7.0, 2000]

If you need to sort by a particular column often, consider making that column a clustered index. This is because the data is already presorted for you and SQL Server is smart enough not to resort the data. [6.5, 7.0, 2000]

If your SELECT statement includes an IN operator along with a list of values to be tested in the query then order the list of values so that the most frequently found values are placed at the first of the list, and the less frequently found values are placed at the end of the list. This can speed performance because the IN option returns true as soon as any of the values in the list produce a match. The sooner the match is made, the faster the query completes. [6.5, 7.0, 2000]

If you need to use the SELECT INTO option then keep in mind that it can lock system tables, preventing others users from accessing the data they need. If you do need to use SELECT INTO, try to schedule it when your SQL Server is less busy, and try to keep the amount of data inserted to a minimum. [6.5, 7.0, 2000]
 
If your SELECT statement contains a HAVING clause then write your query so that the WHERE clause does most of the work (removing undesired rows) instead of the HAVING clause do the work of removing undesired rows. Using the WHERE clause appropriately can eliminate unnecessary rows before they get to the GROUP BY and HAVING clause, saving some unnecessary work, and boosting performance.

For example, in a SELECT statement with WHERE, GROUP BY, and HAVING clauses, here's what happens. First, the WHERE clause is used to select the appropriate rows that need to be grouped. Next, the GROUP BY clause divides the rows into sets of grouped rows, and then aggregates their values. And last, the HAVING clause then eliminates undesired aggregated groups. If the WHERE clause is used to eliminate as many of the undesired rows as possible then this means the GROUP BY and the HAVING clauses will have less work to do, boosting the overall performance of the query. [6.5, 7.0, 2000]

If your application performs many wildcard (LIKE %) text searches on CHAR or VARCHAR columns then consider using SQL Server's full-text search option. The Search Service can significantly speed up wildcard searches of text stored in a database. [7.0, 2000]

The GROUP BY clause can be used with or without an aggregate function. But if you want optimum performance, don't use the GROUP BY clause without an aggregate function. This is because you can accomplish the same end result by using the DISTINCT option instead, and it is faster.

For example, you could write your query in one of two ways:

  1. USE Northwind  
  2. SELECT OrderID  
  3. FROM [Order Details]  
  4. WHERE UnitPrice > 10  
  5. GROUP BY OrderID  
or:
  1. USE Northwind  
  2. SELECT DISTINCT OrderID  
  3. FROM [Order Details]  
  4. WHERE UnitPrice > 10  
Both of the preceding queries produce the same results, but the second one will use less resources and perform faster. [6.5, 7.0, 2000]

The GROUP BY clause can be sped up if you follow this suggestion:

  • Keep the number of rows returned by the query as small as possible.
  • Keep the number of groupings as few as possible.
  • Don't group redundant columns.
  • If there is a JOIN in the same SELECT statement that has a GROUP BY, try to rewrite the query to use a subquery instead of using a JOIN. If this is possible, performance will be faster. If you need to use a JOIN, try to make the GROUP BY column from the same table as the column or columns on which the set function is used.

Consider adding an ORDER BY clause to the SELECT statement that orders by the same column as the GROUP BY. This may cause the GROUP BY to perform faster. Test this to see if is true in your particular situation.
[7.0, 2000]

Sometimes perception is more important than reality. For example, which of the following two queries is the fastest:

  • A query that takes 30 seconds to run, and then displays all of the required results.
  • A query that takes 60 seconds to run, but displays the first screen full of records in less than 1 second.

Most DBAs would choose the first option since it takes less server resources and performs faster. But from many user's point-of-view, the second one may be more palatable. By getting immediate feedback, the user gets the impression that the application is fast, even though in the background, it is not.

If you run into situations where perception is more important than raw performance, consider using the FAST query hint. The FAST query hint is used with the SELECT statement using this form:

OPTION(FAST number_of_rows)

Where number_of_rows is the number of rows that are to be displayed as fast as possible.

When this hint is added to a SELECT statement, it tells the Query Optimizer to return the specified number of rows as fast as possible, without regard to how long it will take to perform the overall query. Before rolling out an application using this hint, I would suggest you test it thoroughly to see that it performs as you expect. You may determine that the query may take about the same amount of time whether the hint is used or not. If this the case, then don't use the hint. [7.0, 2000]

Instead of using temporary tables, consider using a derived table instead. A derived table is the result of using a SELECT statement in the FROM clause of an existing SELECT statement. By using derived tables instead of temporary tables, we can reduce I/O and boost our application's performance. [7.0, 2000]

SQL Server 2000 offers a new data type called "table". Its main purpose is for the temporary storage of a set of rows. A variable, of type "table" behaves as if it is a local variable. And like local variables, it has a limited scope, which is within the batch, function, or Stored Procedure in which it was declared. In most cases, a table variable can be used like a normal table. SELECTs, INSERTs, UPDATEs, and DELETEs can all be made against a table variable.

For best performance, if you need a temporary table in your Transact-SQL code then try to use a table variable instead of creating a conventional temporary table instead. Table variables are created and manipulated in memory instead of the tempdb database, making them much faster. In addition, table variables found in Stored Procedures result in fewer compilations (than when using temporary tables), and transactions using table variables only last as long as the duration of an update on the table variable, requiring less locking and logging resources. [2000]

It is a fairly common request to write a Transact-SQL query to compare a parent table and a child table and determine if there are any parent records that don't have a match in the child table. Generally, there are three ways this can be done.

Using a NOT EXISTS

  1. SELECT a.hdr_key  
  2. FROM hdr_tbl a  
  3. WHERE NOT EXISTS (SELECT * FROM dtl_tbl b WHERE a.hdr_key = b.hdr_key)   
Using a Left Join
  1. SELECT a.hdr_key  
  2. FROM hdr_tbl a  
  3. LEFT JOIN dtl_tbl b ON a.hdr_key = b.hdr_key  
  4. WHERE b.hdr_key IS NULL  
Using a NOT IN
  1. SELECT hdr_key  
  2. FROM hdr_tbl  
  3. WHERE hdr_key NOT IN (SELECT hdr_key FROM dtl_tbl)   
In each case, the preceding query will return identical results. But, which of these three variations of the same query produces the best performance? Assuming everything else is equal, the best performing version through the worst performing version will be from top to bottom, as displayed above. In other words, the NOT EXISTS variation of this query is generally the most efficient. 

I say generally, because the indexes found on the tables, along with the number of rows in each table, can influence the results. If you are not sure which variation to try yourself then you can try them all and see which produces the best results in your particular circumstances. [7.0, 2000]

Be careful when using OR in your WHERE clause, it is fairly simple to accidentally retrieve much more data than you need, that hurts performance. For example, have a look at the query below:

  1. SELECT companyid, plantid, formulaid  
  2. FROM batchrecords  
  3. WHERE companyid = '0001' and plantid = '0202' and formulaid = '39988773'  
or:
  1. companyid = '0001' and plantid = '0202'   
As you can see from this query, the WHERE clause is redundant, since:
  1. companyid = '0001' and plantid = '0202' and formulaid = '39988773'   
Is a subset of:
  1. companyid = '0001' and plantid = '0202'   
In other words, this query is redundant. Unfortunately, the SQL Server Query Optimizer isn't smart enough to know this, and will do exactly what you tell it to. What will happen is that SQL Server will need to retrieve all the data you have requested, then in effect do a SELECT DISTINCT to remove redundant rows it unnecessarily finds. 

In this case, if you drop this code from the query:

or:

companyid = '0001' and plantid = '0202'

Then run the query, you will receive the same results, but with much faster performance. [6.5, 7.0, 2000]

If you need to verify the existence of a record in a table then don't use SELECT COUNT(*) in your Transact-SQL code to identify it, which is very inefficient and wastes server resources. Instead, use the Transact-SQL IF EXITS to determine if the record in question exists, which is much more efficient. For example, here's how you might use COUNT(*):

  1. IF (SELECT COUNT(*) FROM table_name WHERE column_name = 'xxx')  
Here's a faster way, using IF EXISTS:
  1. IF EXISTS (SELECT * FROM table_name WHERE column_name = 'xxx')  
The reason IF EXISTS is faster than COUNT(*) is that the query can end immediately when the text is proven true, while COUNT(*) must count every record, whether there is only one, or thousands, before it can be determined to be true. [7.0, 2000]

Let's say that you often need to INSERT the same value into a column. For example, perhaps you need to perform 100,000 INSERTs a day into a particular table, and that 90% of the time the data inserted into one of the columns of the table is the same value.

If this the case, you can reduce network traffic (along with some SQL Server overhead) by creating this particular column with a default value of the most common value. This way, when you INSERT your data, and the data is the default value, you don't insert any data into this column, instead allowing the default value to be automatically filled in for you. But when the value needs to be different, you will of course INSERT that value into the column. [6.5, 7.0, 2000]

Performing UPDATES takes extra resources for SQL Server to perform. When performing an UPDATE, try to do as many of the following recommendations as you can to reduce the amount of resources required to perform an UPDATE.

The more of the following suggestions you can do, the faster the UPDATE will perform.
  • If you are updating a column of a row that has a unique index then try to only update one row at a time.

  • Try not to change the value of a column that is also the primary key.

  • When updating VARCHAR columns, try to replace the contents with contents of the same length.

  • Try to minimize the updating of tables that have update triggers.

  • Try to avoid updating columns that will be replicated to other databases.

  • Try to avoid updating heavily indexed columns.

  • Try to avoid updating a column that has a reference in the WHERE clause to the column being updated.

Of course, you may have very little choice when updating your data, but at least give the preceding suggestions a thought. [6.5, 7.0, 2000]

If you have created a complex transaction that includes several parts, one part of which has a higher probability of rolling back the transaction than the others then better performance will be provided if you locate the part of the transaction most likely to fail at the front of the greater transaction. This way, if this more-likely-to-fail transaction must be rolled back because of a failure then there has been no resources wasted on the other less-likely-to-fail transactions. [6.5, 7.0, 2000]

Transact-SQL Optimization Tips

  • Try to restrict the queries result set by using the WHERE clause

    This can result in good performance benefits, because SQL Server will return to the client only specific rows, not all rows from the table(s). This can reduce network traffic and boost the overall performance of the query.

  • Try to restrict the queries result set by returning only the particular columns from the table, not all of the table's columns

    This can improve performance because SQL Server will return to the client only specific columns, not all of the table's columns. This can reduce network traffic and boost the overall performance of the query.

  • Use views and Stored Procedures instead of heavy-duty queries

    This can reduce network traffic, because your client will send to the server only a Stored Procedure or view name (perhaps with some parameters) instead of large heavy-duty queries text. This can facilitate permission management also, because you can restrict the user's access to table columns they should not see.

  • Try to avoid using SQL Server cursors, whenever possible

    SQL Server cursors can result in some performance degradation in comparison with select statements. Try to use correlated subquery or derived tables, if you need to perform row-by-row operations.

  • If you need to return the total table's row count, you can use alternative way instead of SELECT COUNT(*) statement

    Because a SELECT COUNT(*) statement makes a full table scan to return the total table's row count, it can require many scans for a large table. There is another way to determine the total row count in a table. You can use the sysindexes system table, in this case. There is a ROWS column in the sysindexes table. This column contains the total row count for each table in your database. So, you can use the following select statement instead of SELECT COUNT(*):
    SELECT rows FROM sysindexes WHERE id = OBJECT_ID('table_name') AND indid < 2
    So you can improve the speed of such queries several times.

  • Try to use constraints instead of triggers, whenever possible

    Constraints are much more efficient than triggers and can boost performance. So, you should use constraints instead of triggers, whenever possible.

  • Use table variables instead of temporary tables

    Table variables require less locking and logging resources than temporary tables, so table variables should be used whenever possible. Table variables are available in SQL Server 2000 only.

  • Try to avoid the HAVING clause, whenever possible

    The HAVING clause can restrict the result set returned by the GROUP BY clause. When you use GROUP BY with the HAVING clause, the GROUP BY clause divides the rows into sets of grouped rows and aggregates their values, and then the HAVING clause eliminates undesired aggregated groups. In many cases, you can write your select statement so, that it will contain only WHERE and GROUP BY clauses without HAVING clause. This can improve the performance of your query.

  • Try to avoid using the DISTINCT clause, whenever possible

    Because using the DISTINCT clause will result in some performance degradation, you should use this clause only when it is necessary.

  • Include a SET NOCOUNT ON statement in your Stored Procedures to stop the message indicating the number of rows affected by a T-SQL statement

    This can reduce network traffic, because your client will not receive the message indicating the number of rows affected by a T-SQL statement.

  • Use select statements with a TOP keyword or the SET ROWCOUNT statement, if you need to return only the first n rows

    This can improve performance of your queries, because the smaller result set will be returned. This can also reduce the traffic between the server and the client.

  • Use the FAST number_rows table hint if you need to quickly return 'number_rows' rows

    You can quickly get the n rows and can work with them, when the query continues execution and produces its full result set.

  • Try to use a UNION ALL statement instead of UNION, whenever possible

    The UNION ALL statement is much faster than UNION, because the UNION ALL statement does not look for duplicate rows, and the UNION statement does look for duplicate rows, whether or not they exist.

  • Do not use optimizer hints in your queries

    Because SQL Server query optimizer is very clever, it is very unlikely that you can optimize your query by using optimizer hints, more often, this will hurt performance.