关于数据库:StoneDB技术解析验证相关数据包是否需要解压缩

25次阅读

共计 8502 个字符，预计需要花费 22 分钟才能阅读完成。

在 StoneDB 中，数据包分为以下几类：

  1. 不相干的数据包：不满足查问条件的数据包。
  2. 相干的数据包：满足查问条件的数据包。
  3. 可疑的数据包：数据包中的数据局部满足查问条件，须要进一步解压缩数据包能力失去满足条件的数据行。

通过对数据包的划分，常识网格技术过滤掉不相干的数据包，读取相干的数据包和可疑的数据包。其中相干的数据包不须要解压缩，只读取元数据，不会产生 IO，可疑的数据包须要解压缩，会产生 IO。

1）创立表 t_user
CREATE TABLE t_user(
  id INT NOT NULL AUTO_INCREMENT,
  first_name VARCHAR(10) NOT NULL,
  last_name VARCHAR(10) NOT NULL,
  sex VARCHAR(5) NOT NULL,
  score INT NOT NULL,
  copy_id INT NOT NULL,
  PRIMARY KEY (`id`),
  key idx_lastname(last_name)
) engine=STONEDB;

2）创立存储过程
DELIMITER //
    create PROCEDURE add_user(in num INT)
    BEGIN
        DECLARE rowid INT DEFAULT 0;
        DECLARE firstname CHAR(1);
        DECLARE name1 CHAR(1);
        DECLARE name2 CHAR(1);
        DECLARE lastname VARCHAR(3) DEFAULT '';
        DECLARE sex CHAR(1);
        DECLARE score CHAR(2);
        WHILE rowid < num DO
        SET firstname = SUBSTRING('赵钱孙李周吴郑王林杨柳刘孙陈江阮侯邹高彭徐',FLOOR(1+21*RAND()),1); 
        SET name1 = SUBSTRING('一二三四五六七八九十甲乙丙丁静景京晶名明铭敏闵民军君俊骏天田甜兲恬益依成城诚立莉力黎励',ROUND(1+43*RAND()),1); 
        SET name2 = SUBSTRING('一二三四五六七八九十甲乙丙丁静景京晶名明铭敏闵民军君俊骏天田甜兲恬益依成城诚立莉力黎励',ROUND(1+43*RAND()),1); 
        SET sex=FLOOR(0 + (RAND() * 2));
        SET score= FLOOR(40 + (RAND() *60));
        SET rowid = rowid + 1;
        IF ROUND(RAND())=0 THEN 
        SET lastname =name1;
        END IF;
        IF ROUND(RAND())=1 THEN
        SET lastname = CONCAT(name1,name2);
        END IF;
        insert INTO t_user(first_name,last_name,sex,score,copy_id) VALUES (firstname,lastname,sex,score,rowid);  
        END WHILE;
    END //
DELIMITER ;

3）插入数据
call add_user(10000000);

4）创立表 t_user_innodb
create table t_user_innodb like t_user;
insert into t_user_innodb select * from t_user;
alter table t_user_innodb engine=innodb;

1）验证读取相干数据包

SQL 的语义逻辑是对字段 first_name 进行分组统计，在 StoneDB 中，元数据信息记录在元数据节点，如果能通过元数据节点读取到元数据，就不须要解压缩数据包，不产生 IO。
在 InnoDB 中，表的统计信息记录在 mysql.innodb_table_stats，优化器依据表和索引的统计信息，生成一个最优的执行打算，而后执行 SQL。别离在 InnoDB 与 StoneDB 执行，通过 SQL profile 察看读取 IO 的状况。
注：为躲避缓存的影响，每组测试前重启数据库实例。

InnoDB

mysql> set profiling=on;                                          
Query OK, 0 rows affected, 1 warning (0.00 sec)

mysql> select first_name,count(*) from t_user_innodb group by first_name;
+------------+----------+
| first_name | count(*) |
+------------+----------+
| 侯         |   476424 |
| 刘         |   475764 |
| 吴         |   475979 |
| 周         |   475891 |
| 孙         |   950444 |
| 彭         |   476632 |
| 徐         |   476219 |
| 李         |   475521 |
| 杨         |   476026 |
| 林         |   477289 |
| 柳         |   476250 |
| 江         |   476623 |
| 王         |   475119 |
| 赵         |   476529 |
| 邹         |   476852 |
| 郑         |   476379 |
| 钱         |   476829 |
| 阮         |   476336 |
| 陈         |   476746 |
| 高         |   476148 |
+------------+----------+
20 rows in set (8.62 sec)

mysql> show profiles;                                                    
+----------+------------+-------------------------------------------------------------------+
| Query_ID | Duration   | Query                                                             |
+----------+------------+-------------------------------------------------------------------+
|        1 | 8.61591075 | select first_name,count(*) from t_user_innodb group by first_name |
+----------+------------+-------------------------------------------------------------------+
1 row in set, 1 warning (0.00 sec)

mysql> show profile cpu,block io for query 1;                            
+----------------------+----------+----------+------------+--------------+---------------+
| Status               | Duration | CPU_user | CPU_system | Block_ops_in | Block_ops_out |
+----------------------+----------+----------+------------+--------------+---------------+
| starting             | 0.000149 | 0.000059 |   0.000083 |            0 |             0 |
| checking permissions | 0.000027 | 0.000011 |   0.000015 |            0 |             0 |
| Opening tables       | 0.048181 | 0.003919 |   0.007952 |          608 |             0 |
| init                 | 0.000036 | 0.000014 |   0.000021 |            0 |             0 |
| System lock          | 0.000022 | 0.000009 |   0.000013 |            0 |             0 |
| optimizing           | 0.000017 | 0.000007 |   0.000010 |            0 |             0 |
| statistics           | 0.000029 | 0.000012 |   0.000016 |            0 |             0 |
| preparing            | 0.000022 | 0.000009 |   0.000013 |            0 |             0 |
| Creating tmp table   | 0.000045 | 0.000019 |   0.000027 |            0 |             0 |
| Sorting result       | 0.000016 | 0.000007 |   0.000009 |            0 |             0 |
| executing            | 0.000014 | 0.000005 |   0.000008 |            0 |             0 |
| Sending data         | 8.566974 | 6.905969 |   0.772964 |       873888 |             0 |
| Creating sort index  | 0.000144 | 0.000164 |   0.000037 |           64 |             0 |
| end                  | 0.000014 | 0.000012 |   0.000003 |           32 |             0 |
| query end            | 0.000028 | 0.000038 |   0.000009 |            0 |             0 |
| removing tmp table   | 0.000019 | 0.000015 |   0.000003 |            0 |             0 |
| query end            | 0.000012 | 0.000010 |   0.000002 |            0 |             0 |
| closing tables       | 0.000031 | 0.000025 |   0.000006 |            0 |             0 |
| freeing items        | 0.000032 | 0.000027 |   0.000006 |            0 |             0 |
| logging slow query   | 0.000067 | 0.000054 |   0.000012 |            0 |             8 |
| cleaning up          | 0.000035 | 0.000028 |   0.000006 |            0 |             0 |
+----------------------+----------+----------+------------+--------------+---------------+
21 rows in set, 1 warning (0.00 sec)

从 SQL profile 可知，SQL 在 InnoDB 执行的过程中，产生 IO 的阶段有 Opening tables、Sending data、Creating sort index、end，其中 Opening tables 是每张表第一次加载都会经验的，可排除探讨。重点关注 Sending data 局部，它示意在执行器的任意阶段，通常是存储引擎层与 Server 层的 IO 交互过程。

StoneDB

mysql> set profiling=on;
Query OK, 0 rows affected, 1 warning (0.00 sec)

mysql> select first_name,count(*) from t_user group by first_name;
+------------+----------+
| first_name | count(*) |
+------------+----------+
| 赵         |   476529 |
| 徐         |   476219 |
| 王         |   475119 |
| 阮         |   476336 |
| 柳         |   476250 |
| 侯         |   476424 |
| 孙         |   950444 |
| 郑         |   476379 |
| 高         |   476148 |
| 林         |   477289 |
| 邹         |   476852 |
| 彭         |   476632 |
| 李         |   475521 |
| 吴         |   475979 |
| 刘         |   475764 |
| 钱         |   476829 |
| 周         |   475891 |
| 杨         |   476026 |
| 陈         |   476746 |
| 江         |   476623 |
+------------+----------+
20 rows in set (0.59 sec)

mysql> show profiles;
+----------+------------+------------------------------------------------------------+
| Query_ID | Duration   | Query                                                      |
+----------+------------+------------------------------------------------------------+
|        1 | 0.59069975 | select first_name,count(*) from t_user group by first_name |
+----------+------------+------------------------------------------------------------+
1 row in set, 1 warning (0.00 sec)

mysql> show profile cpu,block io for query 1;
+----------------------+----------+----------+------------+--------------+---------------+
| Status               | Duration | CPU_user | CPU_system | Block_ops_in | Block_ops_out |
+----------------------+----------+----------+------------+--------------+---------------+
| starting             | 0.000160 | 0.000066 |   0.000089 |            0 |             0 |
| checking permissions | 0.000027 | 0.000011 |   0.000015 |            0 |             0 |
| Opening tables       | 0.011405 | 0.003718 |   0.007688 |            0 |           240 |
| System lock          | 0.000385 | 0.000163 |   0.000222 |            0 |             0 |
| init                 | 0.000050 | 0.000021 |   0.000028 |            0 |             0 |
| optimizing           | 0.000143 | 0.000061 |   0.000082 |            0 |             0 |
| update multi-index   | 0.000052 | 0.000022 |   0.000030 |            0 |             0 |
| aggregation          | 0.578315 | 2.639504 |   0.981471 |            0 |             8 |
| query end            | 0.000069 | 0.000043 |   0.000026 |            0 |             0 |
| closing tables       | 0.000035 | 0.000021 |   0.000013 |            0 |             0 |
| freeing items        | 0.000034 | 0.000021 |   0.000013 |            0 |             0 |
| cleaning up          | 0.000027 | 0.000017 |   0.000010 |            0 |             0 |
+----------------------+----------+----------+------------+--------------+---------------+
12 rows in set, 1 warning (0.00 sec)

从 SQL profile 可知，SQL 在 StoneDB 执行的过程中，只在 Opening tables 阶段产生 IO。其它阶段没有产生 IO，阐明相干数据包是不须要解压缩的，通过元数据失去。

2）验证读取可疑数据包

SQL 的语义逻辑是查问一行数据，StoneDB 能够通过常识网格技术过滤掉不相干的数据包，因为只返回一行数据，最终只能找到可疑的数据包，而后解压缩可疑的数据包，最终失去这一行数据。InnoDB 还是依据统计信息生成一个最优的执行打算去执行 SQL。

InnoDB

mysql> set profiling=on;                              
Query OK, 0 rows affected, 1 warning (0.00 sec)

mysql> select count(*) from t_user_innodb where first_name='柳' and copy_id=9968888;
+----------+
| count(*) |
+----------+
|        1 |
+----------+
1 row in set (3.20 sec)

mysql> show profile cpu,block io for query 1;
+----------------------+----------+----------+------------+--------------+---------------+
| Status               | Duration | CPU_user | CPU_system | Block_ops_in | Block_ops_out |
+----------------------+----------+----------+------------+--------------+---------------+
| starting             | 0.000170 | 0.000072 |   0.000092 |            0 |             0 |
| checking permissions | 0.000030 | 0.000013 |   0.000016 |            0 |             0 |
| Opening tables       | 0.024121 | 0.004351 |   0.008638 |          800 |             0 |
| init                 | 0.000049 | 0.000021 |   0.000027 |            0 |             0 |
| System lock          | 0.000019 | 0.000008 |   0.000011 |            0 |             0 |
| optimizing           | 0.000022 | 0.000010 |   0.000012 |            0 |             0 |
| statistics           | 0.000030 | 0.000013 |   0.000016 |            0 |             0 |
| preparing            | 0.000026 | 0.000012 |   0.000015 |            0 |             0 |
| executing            | 0.000013 | 0.000005 |   0.000007 |            0 |             0 |
| Sending data         | 3.169882 | 2.755171 |   0.389367 |       534176 |             0 |
| end                  | 0.000069 | 0.000050 |   0.000018 |            0 |             0 |
| query end            | 0.000029 | 0.000022 |   0.000008 |            0 |             0 |
| closing tables       | 0.000031 | 0.000023 |   0.000009 |            0 |             0 |
| freeing items        | 0.000035 | 0.000025 |   0.000009 |            0 |             0 |
| cleaning up          | 0.000038 | 0.000028 |   0.000010 |            0 |             0 |
+----------------------+----------+----------+------------+--------------+---------------+
15 rows in set, 1 warning (0.00 sec)

StoneDB

mysql> set profiling=on;                                                            
Query OK, 0 rows affected, 1 warning (0.00 sec)

mysql> select count(*) from t_user where first_name='柳' and copy_id=9968888;
+----------+
| count(*) |
+----------+
|        1 |
+----------+
1 row in set (0.01 sec)

mysql> show profile cpu,block io for query 1;                                
+----------------------+----------+----------+------------+--------------+---------------+
| Status               | Duration | CPU_user | CPU_system | Block_ops_in | Block_ops_out |
+----------------------+----------+----------+------------+--------------+---------------+
| starting             | 0.000173 | 0.000081 |   0.000086 |            0 |             0 |
| checking permissions | 0.000026 | 0.000013 |   0.000013 |            0 |             0 |
| Opening tables       | 0.010228 | 0.009385 |   0.000843 |            0 |           240 |
| System lock          | 0.000232 | 0.000113 |   0.000119 |            0 |             0 |
| init                 | 0.000045 | 0.000021 |   0.000022 |            0 |             0 |
| optimizing           | 0.000144 | 0.000071 |   0.000074 |            0 |             0 |
| update multi-index   | 0.003694 | 0.002027 |   0.006428 |            0 |             0 |
| aggregation          | 0.000191 | 0.000093 |   0.000098 |            0 |            16 |
| query end            | 0.000020 | 0.000010 |   0.000010 |            0 |             0 |
| closing tables       | 0.000029 | 0.000014 |   0.000015 |            0 |             0 |
| freeing items        | 0.000033 | 0.000016 |   0.000017 |            0 |             0 |
| cleaning up          | 0.000027 | 0.000013 |   0.000013 |            0 |             0 |
+----------------------+----------+----------+------------+--------------+---------------+
12 rows in set, 1 warning (0.00 sec)

从 SQL profile 可知，SQL 在 StoneDB 执行的过程中，在 aggregation 阶段产生 IO。
综上所述：

  1. 常识网格技术过滤出相干的数据包后，只须要读取元数据，不再解压缩数据包；
  2. 常识网格技术过滤出可疑的数据包后，须要解压缩数据包。

正文完
数据库
发表至： 数据库
2022-08-16
 0
关于数据库:数据库存储与索引技术一存储模型与索引结构演变
快讯Apache-ShardingSphere进入CNCF全景图
关于数据库:StarRocks市场渗透率跻身Top10
关于数据库:数仓出现wait-in-ccn-queue的时候怎么迅速定位处理