Hi, Summary tables in particular have this problem. A summary table may have a very long (in terms of bytes) unique key or primary key, but this causes a huge problem for secondary keys. For example: create table summary_table( summary_year int, summary_month int, summary_day int, summary_customer_id bigint, summary_customer_item_category int, summary_cnt bigint, summary_sum decimal(15,5), summary_min decimal(15,5) primary key( summary_year year, summary_month int, summary_day int, summary_customer_id bigint, summary_customer_item_category int ) , key (summary_customer_id), key (summary_customer_item_category, summary_customer_id) ) This table will be clustered on a 28 BYTE primary key. This is not efficient. In this case, I want this table to have an internally clustered key. The tables are maintained by software, and it doesn't make sense to maintain auto_increment values on a summary table. That is ugly and not efficient. There are tons of other examples that I can think of where a long PK is desirable but clustering on the PK is not. I do not want to use a storage engine other than InnoDB. Thank you for considering my feature request. On Wed, Oct 13, 2010 at 11:14 AM, Jobin Augustine