A question often comes when talking about indexing: should we use multiple column indexes or multiple indexes on single columns? Peter Zaitsev wrote about it back in 2008 and the conclusion then was that a multiple column index is most often the best solution. But with all the recent optimizer improvements, is there anything different with MySQL 5.6?

Setup

For this test, we will use these 2 tables (same structure as in Peter’s post):

CREATE TABLE t1000merge (
  id int not null auto_increment primary key,
  i int(11) NOT NULL,
  j int(11) NOT NULL,
  val char(10) NOT NULL,
  KEY i (i),
  KEY j (j)
) ENGINE=InnoDB;
CREATE TABLE t1000idx2 (
  id int not null auto_increment primary key,
  i int(11) NOT NULL,
  j int(11) NOT NULL,
  val char(10) NOT NULL,
  KEY ij (i,j)
) ENGINE=InnoDB;

Tables were …