TiDB Cloud 上 Tiflash 列存功能初体验

简介：

相比于行存，TiFlash 根据强 Schema 按列式存储结构化数据，借助 *** 的向量化计算引擎，带来读取和计算双重性能优势。相较于普通列存，TiFlash 则具有实时更新、分布式自动扩展、SI（Snapshot Isolation）隔离级别读取等优势。

一：TiFlash 对 OLAP 查询加速

OLAP 类的查询通常具有以下几个特点：

· 每次查询读取大量的行，但是仅需要少量的列

· 宽表，即每个表包含着大量的列

· 查询通过一张或多张小表关联一张大表，并对大表上的列做聚合

二：TiFlash 列存引擎针对这类查询有较好的优化效果：

(1) I/O 优化

· 每次查询可以只读取需要的列，减少了 I/O 资源的使用

· 同列数据类型相同，相较于行存可以获得更高的压缩比

· 整体的 I/O 减少，令内存的使用更加高效

(2) CPU 优化

· 列式存储可以很方便地按批处理字段，充分利用 CPU Cache 取得更好的局部性

· 利用向量化处理指令并行处理部分计算

三：实验过程：

（1）创建tidb cloud账号

（2）创建tidb 集群

（3）通过tidb cloud import导入1000条和100万条基础测试数据。

（4）创建客户、订单、产品表，插入测试数据

mysql> CREATE TABLE customers (

    ->     id INT AUTO_INCREMENT PRIMARY KEY,

    ->     name VARCHAR(100),

    ->     email VARCHAR(100),

    ->     created_at DATETIME

    -> ) ;

Query OK, 0 rows affected (0.71 sec)

mysql> CREATE TABLE orders (

    ->     id INT AUTO_INCREMENT PRIMARY KEY,

    ->     customer_id INT,

    ->     product_id INT,

    ->     quantity INT,

    ->     price DECIMAL(10, 2),

    ->     timestamp DATETIME

    -> ) ;

Query OK, 0 rows affected (0.46 sec)

mysql> CREATE TABLE products (

    ->     id INT AUTO_INCREMENT PRIMARY KEY,

    ->     name VARCHAR(100),

    ->     category VARCHAR(100),

    ->     price DECIMAL(10, 2)

    -> ) ;

Query OK, 0 rows affected (0.56 sec)

-- 顾客表插入100万行数据

mysql> INSERT INTO customers (name, email, created_at)

    -> SELECT 

    ->     CONCAT(Customer, seq),

    ->     CONCAT(customer, seq, @example.com),

    ->     NOW()

    -> FROM 

    ->     (SELECT seq FROM seq_1_to_1000000) s;

Query OK, 1000000 rows affected (11.96 sec)

Records: 1000000  Duplicates: 0  Warnings: 0

-- 产品表插入1000行数据

mysql> INSERT INTO products (name, category, price)

    -> SELECT 

    ->     CONCAT(Product, seq),

    ->     CONCAT(Category, FLOOR(RAND()*(10-1+1))+1),

    ->     ROUND(RAND()*(1000-1+1), 2)

    -> FROM 

    ->     (SELECT seq FROM seq_1_to_1000) s;

Query OK, 1000 rows affected (0.91 sec)

Records: 1000  Duplicates: 0  Warnings: 0

mysql> 

（5）通过上图可以看出数据已入库，并且按照行存储。并执行关联，求和操作，耗时1229ms。

（6）构建TIFLASH 副本

mysql> ALTER TABLE test.customers SET TIFLASH REPLICA 1;

Query OK, 0 rows affected, 1 warning (0.87 sec)

mysql> ALTER TABLE test.orders SET TIFLASH REPLICA 1;

Query OK, 0 rows affected, 1 warning (1.65 sec)

mysql> ALTER TABLE test.products SET TIFLASH REPLICA 1;

Query OK, 0 rows affected, 1 warning (0.85 sec)

查看表同步进度

mysql>  SELECT * FROM information_schema.tiflash_replica WHERE TABLE_SCHEMA = test and TABLE_NAME = customers;

+--------------+------------+----------+---------------+-----------------+-----------+----------+

| TABLE_SCHEMA | TABLE_NAME | TABLE_ID | REPLICA_COUNT | LOCATION_LABELS | AVAILABLE | PROGRESS |

+--------------+------------+----------+---------------+-----------------+-----------+----------+

| test         | customers  |      123 |             2 |                 |         0 |        0 |

+--------------+------------+----------+---------------+-----------------+-----------+----------+

1 row in set (0.64 sec)

mysql> SELECT * FROM information_schema.tiflash_replica WHERE TABLE_SCHEMA = test and TABLE_NAME = orders;

+--------------+------------+----------+---------------+-----------------+-----------+----------+

| TABLE_SCHEMA | TABLE_NAME | TABLE_ID | REPLICA_COUNT | LOCATION_LABELS | AVAILABLE | PROGRESS |

+--------------+------------+----------+---------------+-----------------+-----------+----------+

| test         | orders     |      134 |             2 |                 |         0 |        0 |

+--------------+------------+----------+---------------+-----------------+-----------+----------+

1 row in set (0.49 sec)

mysql> SELECT * FROM information_schema.tiflash_replica WHERE TABLE_SCHEMA = test and TABLE_NAME = products;

+--------------+------------+----------+---------------+-----------------+-----------+----------+

| TABLE_SCHEMA | TABLE_NAME | TABLE_ID | REPLICA_COUNT | LOCATION_LABELS | AVAILABLE | PROGRESS |

+--------------+------------+----------+---------------+-----------------+-----------+----------+

| test         | products   |      128 |             2 |                 |         0 |        0 |

+--------------+------------+----------+---------------+-----------------+-----------+----------+

1 row in set (0.52 sec)

AVAILABLE 字段表示该表的 TiFlash 副本是否可用。1 代表可用，0 代表不可用。副本状态变为可用之后就不再改变，如果通过 DDL 命令修改副本数则会重新计算同步进度。

PROGRESS 字段代表同步进度，在 0.0~1.0 之间，1 代表至少 1 个副本已经完成同步。刚才查询lineitem表副本不可用，再次查询已经可用

mysql>  SELECT * FROM information_schema.tiflash_replica WHERE TABLE_SCHEMA = test and TABLE_NAME = customers;

+--------------+------------+----------+---------------+-----------------+-----------+----------+

| TABLE_SCHEMA | TABLE_NAME | TABLE_ID | REPLICA_COUNT | LOCATION_LABELS | AVAILABLE | PROGRESS |

+--------------+------------+----------+---------------+-----------------+-----------+----------+

| test         | customers  |      123 |             2 |                 |         1 |        1 |

+--------------+------------+----------+---------------+-----------------+-----------+----------+

1 row in set (1.94 sec)

mysql> SELECT * FROM information_schema.tiflash_replica WHERE TABLE_SCHEMA = test and TABLE_NAME = products;

+--------------+------------+----------+---------------+-----------------+-----------+----------+

| TABLE_SCHEMA | TABLE_NAME | TABLE_ID | REPLICA_COUNT | LOCATION_LABELS | AVAILABLE | PROGRESS |

+--------------+------------+----------+---------------+-----------------+-----------+----------+

| test         | products   |      128 |             2 |                 |         1 |        1 |

+--------------+------------+----------+---------------+-----------------+-----------+----------+

1 row in set (0.77 sec)

mysql> SELECT * FROM information_schema.tiflash_replica WHERE TABLE_SCHEMA = test and TABLE_NAME = orders;

+--------------+------------+----------+---------------+-----------------+-----------+----------+

| TABLE_SCHEMA | TABLE_NAME | TABLE_ID | REPLICA_COUNT | LOCATION_LABELS | AVAILABLE | PROGRESS |

+--------------+------------+----------+---------------+-----------------+-----------+----------+

| test         | orders     |      134 |             2 |                 |         1 |        1 |

+--------------+------------+----------+---------------+-----------------+-----------+----------+

1 row in set (0.66 sec)

（7）如图所示，发送进行了列存储。执行相同的SQL查询。

结论：如图所示相同SQL，执行计划使用了TiFlash ，执行时间118ms；相比未使用TiFlash 执行时间1229ms，速度提升了10倍。

备注：一条select 查询中，可以同时使用TIKV和TIFLSH，tidb默认会根据select查询进行判断。