MySQL 磁盘 I/O 当 select 执行有效时
MySQL disk I/O when select execution works
我对mysql的底层原理有疑问。
假设我有一个 table 有 4 列并且没有索引。
select * 来自 A 与 select 第一列来自 A.
那么读取I/O作为性能有区别吗？不是网络 I/O.
很好奇从磁盘读取数据的原理。
In MySQL's default storage engine, InnoDB, the whole row is stored together. So there's little difference between SELECT * and SELECT first_column. It still has to do the I/O to load the page that contains each row, and once it has done so, it has all the columns in the buffer pool, so there's no significant increase in cost to reading more columns.
There's an exception case: if you have long strings in VARCHAR, VARBINARY, TEXT, or BLOB columns, the part of the string that doesn't fit on the page with the row is stored on additional pages elsewhere in the database storage (see https://www.percona.com/blog/2010/02/09/blob-storage-in-innodb/). If you omit string columns you don't need, you can reduce I/O because InnoDB won't need to read the extra pages for the long string data.
The above assumes, as you stated in your question, that there are no indexes. If you do have indexes, you might benefit from the covering index effect: if you reference only the column(s) that are part of the index, InnoDB can avoid reading the page with the whole排。 It will just read the page that contains the referenced index entries. That will help reduce I/O.
我对mysql的底层原理有疑问。
假设我有一个 table 有 4 列并且没有索引。
select * 来自 A 与 select 第一列来自 A.
那么读取I/O作为性能有区别吗？不是网络 I/O.
很好奇从磁盘读取数据的原理。
In MySQL's default storage engine, InnoDB, the whole row is stored together. So there's little difference between SELECT * and SELECT first_column. It still has to do the I/O to load the page that contains each row, and once it has done so, it has all the columns in the buffer pool, so there's no significant increase in cost to reading more columns.
There's an exception case: if you have long strings in VARCHAR, VARBINARY, TEXT, or BLOB columns, the part of the string that doesn't fit on the page with the row is stored on additional pages elsewhere in the database storage (see https://www.percona.com/blog/2010/02/09/blob-storage-in-innodb/). If you omit string columns you don't need, you can reduce I/O because InnoDB won't need to read the extra pages for the long string data.
The above assumes, as you stated in your question, that there are no indexes. If you do have indexes, you might benefit from the covering index effect: if you reference only the column(s) that are part of the index, InnoDB can avoid reading the page with the whole排。 It will just read the page that contains the referenced index entries. That will help reduce I/O.