玩Oracle的都比较关注shared pool，特别是library cache，在使用了绑定变量（预编译sql）之后确实能得到很大的性能提升。现在在转Mysql之后特别是innodb很多东西都还能和Oracle对得上号的，就像innodb_buffer_pool_size类似于Oracle的database buffer cache，innodb_log_buffer_size类似于redo log buffer，但是innodb_additional_mem_pool_size仅仅类似于shared pool的Data dictionary cache，似乎还缺少和library cache相对应的东西。那就有一个问题了，在Mysql里面使用预编译的sql还会有性能提升吗？
这里我用Java的jdbc做了一下测试，分别用Statement和PreparedStatement执行1000个sql，并运行10次
1.使用Statement做硬解析：
1 package exmysql; 2 3 import java.sql.Connection; 4 import java.sql.DriverManager; 5 import java.sql.ResultSet; 6 import java.sql.SQLException; 7 import java.sql.Statement; 8 import java.util.Date; 9 10 public class adddata { 11 12 private static long worker(){ 13 Date begin = new Date(); 14 15 String driver="com.mysql.jdbc.Driver"; 16 17 String url="jdbc:mysql://172.16.2.7:3306/testdb"; 18 19 Connection conn=null; 20 Statement stmt=null; 21 ResultSet rs=null; 22 23 try{ 24 Class.forName(driver); 25 conn=DriverManager.getConnection(url,"dbaadmin","123456"); 26 stmt=conn.createStatement(); 27 String sql; 28 for (int i=1;i<=5000;i++){ 29 sql="select * from test1 where >i; 30 rs=stmt.executeQuery(sql); 31 } 32 } 33 catch(SQLException | ClassNotFoundException e){ 34 e.printStackTrace(); 35 } 36 37 if(stmt!=null){ 38 try{ 39 stmt.close(); 40 } 41 catch(SQLException e){ 42 e.printStackTrace(); 43 } 44 } 45 46 if(conn!=null){ 47 try{ 48 conn.close(); 49 } 50 catch(SQLException e){ 51 e.printStackTrace(); 52 } 53 } 54 55 Date end = new Date(); 56 return end.getTime()-begin.getTime(); 57 } 58 59 public static void main(String[] args) { 60 // TODO Auto-generated method stub 61 62 long elapsed,average; 63 average=0; 64 for (int i=1;i<=10;i++){ 65 elapsed=worker(); 66 System.out.println("elapsed time(ms):"+elapsed); 67 average=average+elapsed; 68 } 69 System.out.println("average time(ms):"+average/10); 70 } 71 72 }
结果如下：
elapsed time(ms):24652
elapsed time(ms):13380
elapsed time(ms):13250
elapsed time(ms):13877
elapsed time(ms):13275
elapsed time(ms):13193
elapsed time(ms):19022
elapsed time(ms):13558
elapsed time(ms):14138
elapsed time(ms):13364
average time(ms):15170
2.同样的sql用PreparedStatement预编译执行
1 package exmysql; 2 3 import java.sql.*; 4 import java.util.Date; 5 6 public class insert_data { 7 8 private static long worker(){ 9 Date begin = new Date(); 10 11 String driver="com.mysql.jdbc.Driver"; 12 13 String url="jdbc:mysql://172.16.2.7:3306/testdb"; 14 15 Connection conn=null; 16 PreparedStatement pstm=null; 17 ResultSet rs=null; 18 19 try{ 20 Class.forName(driver); 21 conn=DriverManager.getConnection(url,"dbaadmin","123456"); 22 String sql="select * from test1 where id=?"; 24 pstm=conn.prepareStatement(sql); 25 for(int i=1;i<=5000;i++){ 26 pstm.setInt(1, i); 27 rs=pstm.executeQuery(); 28 } 29 31 } 32 catch(SQLException | ClassNotFoundException e){ 33 e.printStackTrace(); 34 } 35 36 if(pstm!=null){ 37 try{ 38 pstm.close(); 39 } 40 catch(SQLException e){ 41 e.printStackTrace(); 42 } 43 } 44 45 if(conn!=null){ 46 try{ 47 conn.close(); 48 } 49 catch(SQLException e){ 50 e.printStackTrace(); 51 } 52 } 53 54 Date end = new Date(); 55 return end.getTime()-begin.getTime(); 56 } 57 58 public static void main(String[] args) { 59 // TODO Auto-generated method stub 60 61 long elapsed,average; 62 average=0; 63 for (int i=1;i<=10;i++){ 64 elapsed=worker(); 65 System.out.println("elapsed time(ms):"+elapsed); 66 average=average+elapsed; 67 } 68 System.out.println("average time(ms):"+average/10); 69 } 70 71 }
结果如下：
elapsed time(ms):14773
elapsed time(ms):16352
elapsed time(ms):14797
elapsed time(ms):15800
elapsed time(ms):12069
elapsed time(ms):14953
elapsed time(ms):13238
elapsed time(ms):12366
elapsed time(ms):15263
elapsed time(ms):13089
average time(ms):14270
可以看出两种方式执行的结果几乎相同，不像Oracle差距那么大。而且就算是用PreparedStatement的方式，在Mysql数据库端抓出来的sql语句也不是以变量id=?的形式出现的，而是实际的数值。后来在网上看到在连接字符串上加上useServerPrepStmts=true可以实现真正的预编译
String url="jdbc:mysql://172.16.2.7:3306/testdb"; url=url+"?useServerPrepStmts=true";
加上这段后可以在数据库端可以看到明确的结果：
mysql> show global status like 'Com_stmt_prepare';
+------------------+-------+
| Variable_name | Value |
+------------------+-------+
| Com_stmt_prepare | 11 |
+------------------+-------+
1 row in set (0.00 sec)
但是实际的运行结果和上面几乎一样，性能上也没有任何的提升。由此可以推断出Mysql由于缺少类似于Oracle的library cache的部件，因此采用预编译方式执行sql是没有性能上的提升的。
