Execution plan of MYSQL and TiDB

foreword

The data volume of a tpc-h database is used here to compare query plans. And analyze the execution plan with the help of 22 query statements in tpc-h.

MySQL uses a standard installation, and TiDB uses a stand-alone beta version. The performance results here cannot explain the difference in performance.

The main purpose of this article is to compare the differences between Mysql and TiDB when executing SQL queries.

mysql version 5.7 TiDB version v2.0.0-rc.4

 

Preparation Phase

After data is imported into TiDB, statistics are missing:

SHOW STATS_META

 

Statistics can be refreshed manually

ANALYZE TABLE nation,region,part,supplier,partsupp,customer,orders,lineitem

Check SHOW STATS_META again after refreshing

 

First select Q17 as an example to query

select
    sum(l_extendedprice) / 7.0 as avg_yearly
from
    lineitem,
    part
where 
    p_partkey = l_partkey
     and p_brand =  ' Brand#23 ' # Specify the brand. BRAND='Brand#MN', M and N are two letters, representing two values, independent of each other, the value is between 1 and 5,
     and p_container =  ' MED BOX ' # Specify the packaging type. Randomly select within the range specified by the
     TPC - H standard and l_quantity < (
         select 0.2 * avg (l_quantity)
         from
              
            lineitem
        where
            l_partkey = p_partkey
    );

The part table is 200,000 P_PARTKEY is the primary key, and the lineitem is 6 million

The query time of mysql is about 1 second, and the query time of TiDB is about 30 seconds. By comparison, it is found that mysql will automatically create an index when creating constraints.

 

mysql execution plan:

 mysql first queries the part table, and the 200,000 data has been filtered to several hundred after processing where. Then associate with the lineitem, and finally process the subquery.

The primary key association is borrowed during the query, so

 

TiDB execution plan