如下SQL的优化： select count(*) from ( select id,name,col1,col2 from t1  where name='xxxx' union select id,name ,null as col1,'' as col2 from t2  where id!=1 ) 原SQL慢原因：由于union需要过滤 重复记录。所以会有后台的一个group by id,name,col1,col2的操作。 优化：去掉group by。统计出 t1表，t2 表。去除重复记录的次数 优化后： select sum(a)-sum(dis) from ( SELECT  count(*) as a ,0 as dis  FROM t1 a where name='xxxx' union all SELECT  count(*) as a,0 as dis  FROM  t2 b WHERE status !=8 union all select 0 as a, count(1) as dis from  t1 a inner join t2 b a.id=b.id and a.name=b.name where a.name='xxxx' and a.col1 is null and a.col2='' and b.status!=8  ) as v