1、explain
To have a better understanding of the execution plan, you need to have a simple understanding of the basic structure of MySQL and the basic principles of query. The functional architecture of MySQL itself is divided into three parts, namely the application layer, logical layer, and physical layer. Not only MySQL, most other database products are divided according to this architecture.
The application layer is mainly responsible for interacting with the client, establishing a link, remembering the link status, returning data, and responding to requests. This layer deals with the client.
The logic layer is mainly responsible for query processing, transaction management and other database function processing
Take query as an example.
First, after receiving the query SQL, the database will immediately allocate a thread to process it. The first step is to optimize the SQL query by the query processor. After optimization, an execution plan will be generated and then handed over to the plan executor for execution.
The plan executor needs to access the lower-level transaction manager and storage manager to operate data. Their respective division of labor is different. Finally, the query structure information is obtained by calling the file of the physical layer, and the final result is responded to the application layer.
The physical layer, the files stored on the actual physical disk, mainly includes sub-text data files and log files.
Through the above description, generating an execution plan is an essential step to execute a SQL. The performance of a SQL can be seen intuitively by viewing the execution plan. The execution plan provides various query types and levels, which is convenient for us to perform View and serve as the basis for performance analysis.
1.1, identity column
id indicates the serial number of the query statement, which is automatically allocated and incremented in order. The larger the value, the higher the execution priority.
When the IDs are the same, the priority is from top to bottom.
1.2, select_type column
select_type indicates the query type, common ones include SIMPLE simple query, PRIMARY main query, SUBQUERY subquery, UNION joint query, UNION
RESULT joint temporary table results, etc.
1.3, table columns
table indicates the table name, table alias, and temporary table name queried by the SQL statement.
1.4, partitions column
partitions indicates the partitions matched by the SQL query, and NULL is displayed if there is no partition.
1.5, type column
type indicates the table connection type or data access type, that is, how to establish connection between tables, or how to access data. Specifically, there are the following values, and the order of performance from good to bad is: system > const > eq_ref > ref > fulltext > ref_or_null > index_merge > unique_subquery > index_subquery > range > index > ALL system When there is only one row of records in the table, that is, the system table,
it
is A special column of const type.
const
indicates that the primary key or unique index is used for equivalent query, and at most one record is returned. Good performance, recommended to use.
eq_ref
indicates that the table connection uses the primary key or unique index, and the following SQL uses the primary key id of the user table.
ref
means to use non-unique index for equivalent query.
ref_or_null
indicates that a non-unique index is used for equivalent query and contains rows with null values.
index_merge
indicates the optimization logic used for index merging, that is, multiple indexes used.
range
indicates that the index range query is used.
index
means to use the index for full table scan.
ALL
means a full table scan, with the worst performance.
1.6, possible_keys column
Indicates the index columns that may be used, but may not be used in actual queries.
1.7, key column
Indicates that the actual query uses the index column.
1.8、key_len列
Indicates the number of bytes occupied by the index.
The number of bytes occupied by each type is as follows:
1.9, ref column
Indicates the parameters compared with the index in the where statement or table connection. The common ones are const (constant), func (function), and field name. Displays NULL if no index is used.
1.10. rows
Indicates the number of rows scanned by executing the SQL statement.
1.11. filtered columns
Indicates the percentage of table rows filtered by the condition.
Used to estimate the number of rows scanned when connecting with other tables, row x filtered = 252004 x 10% = 250,000 rows
1.12, Extra column
Indicates some additional extended information, which is not suitable for display in other columns, but is very important.
Using where
indicates that the where conditional search is used, but the index is not used.
Using index
means that the covering index is used, that is, the required data can be found on the index, and there is no need to query the table twice, and the performance is better.
Using filesort
indicates that external sorting is used, that is, the sorting field does not use an index.
Using temporary
means that a temporary table is used. In the following example, a temporary table is used to store query results.
Using join buffer
indicates that when performing table association, no index is used, and the connection buffer is used to store temporary results.
In the following example, user_id is not indexed in both tables.
Using index condition
indicates that the optimization feature of index pushdown is used.
2. Use Profiling for query optimization
MySQL's Query Profiler is a very convenient Query diagnostic analysis tool. Through this tool, you can obtain the consumption of various resources during the entire execution process of a Query, such as CPU, IO, IPC, SWAP, etc., as well as the PAGE FAULTS that occurred. CONTEXT SWITCHE, etc., and at the same time, the position in the source file of each function called by MySQL during the execution of the Query can also be obtained.
Let's take a look at the specific usage of Query Profiler.
Other techniques for query optimization
Index design relies on predicate expressions (conditional expressions or truth expressions----the statement after where)
to avoid
appropriate redundancy of Join data
