Re: Wrong estimation of Rows in explain

[Rick James]

The PARTITIONing provides no performance gain.

The query is asking for 3 million rows from one table, then randomly reaching into the other table 3 million times. Since `innodb_buffer_pool_size` is very small (it should be about 70% of _available_ RAM), the query will have to do a lot of I/O.

I don't see a row estimate of 2; I do see 17. Either of those could be reasonable, since it it not the number of rows in the table, but rather the number of rows in the second table for each row of the first table.

It is usually a mistake to have a pair of columns like this in some table other than a table with "only" these:

AttributeId int(11) ,
AttributeName varchar(100) ,

Do you need 2 billion values for those various IDs? Consider `SMALLINT UNSIGNED` (65K distinct values), or other sizes of INTs. That will save some space.