Re: InnoDB Clustered Indexes

[James Day]

No, it won't achieve your objective. Consider what happens when the index node containing values around the value you're changing to is already full. InnoDB will need to split that index node to make room and you'll end up with two partially filled nodes.

The source code for the InnoDB engine will tell you exactly how the data is laid out on disk.

I discussed the fragmentation issue with Heikki last week, wondering if there was some possibility of background node reorganization to remove most of the benefit from doing periodic OPTIMIZE TABLE commands. He seemed receptive but that's not something which means it'll happen at any predictable time in the future, unless it's important enough to a customer that they want to sponsor the feature.